Journal of Personality and Social Psychology 2004, Vol. 86, No.

1, 112129

Copyright 2004 by the American Psychological Association, Inc. 0022-3514/04/$12.00 DOI: 10.1037/0022-3514.86.1.112

Intelligence: Genetics, Genes, and Genomics

Robert Plomin and Frank M. Spinath
Kings College London
More is known about the genetics of intelligence than about any other trait, behavioral or biological, which is selectively reviewed in this article. Two of the most interesting genetic findings are that heritability of intelligence increases throughout the life span and that the same genes affect diverse cognitive abilities. The most exciting direction for genetic research on intelligence is to harness the power of the Human Genome Project to identify some of the specific genes responsible for the heritability of intelligence. The next research direction will be functional genomicsfor example, understanding the brain pathways between genes and intelligence. Deoxyribonucleic acid (DNA) will integrate life sciences research on intelligence; bottom-up molecular biological research will meet top-down psychological research in the brain.

As indicated in the preface to this special section on intelligence, the centenary of Spearmans seminal article on intelligence (Spearman, 1904) is an appropriate moment to take stock of what we know about this oldest of personological constructs at diverse levels of analysis genes, physiology, psychology, and sociologythroughout the life span. More is known about the genetics of individual differences in intelligence than any other behavioral trait. This research is reviewed briefly in the first section of the present articleGenetics. The second sectionGenes describes current attempts to harness the power of the Human Genome Project in order to identify some of the presumably many genes responsible for the heritability of intelligence. The third section Genomics discusses the next step, functional genomics, which attempts to chart pathways between genes and intelligence. To be able to address these issues, this article needs to assume basic understanding of the psychometric construct of intelligence, which is described in the preface and in other articles in this special section (see also books by Bock, Goode, & Webb, 2000; Brody, 1992; Deary, 2000). The problem with the word intelligence is that it means different things to different people. The present article uses the psychometric definition of intelligence as general cognitive abilitySpearmans g, which was discovered at the same time that Mendels laws of inheritance were rediscovered (Spearman, 1904). That is, the word intelligence will be used to refer to the substantial covariation among diverse measures of cognitive ability as indexed by an unrotated first principalcomponent score, which typically accounts for about 40% of the total variance of diverse cognitive tests, or by a total score across diverse tests as is done in intelligence tests (Jensen, 1998). We also

forgo a discussion of the relationship between intelligence and personality, which is discussed in the preface to this special section. In terms of genetics, reviews have been written about traditional personality traits such as the five-factor model (Loehlin, 1992; Plomin, DeFries, McClearn, & McGuffin, 2001); intelligence is especially related to Openness to Experience (McCrae & Costa, 1997). Personality and its relationship to psychopathology has also increasingly become a target for molecular genetic research aimed at identifying specific genes responsible for the ubiquitous heritability of personality traits (Benjamin, Ebstein, & Belmaker, 2002). It is also necessary to assume at least some passing familiarity with quantitative genetics and molecular genetics. Quantitative genetics is a theory of familial resemblance for complex traits that leads to methods like the twin method and adoption method, which decompose phenotypic variance into genetic and environmental components of variance. Molecular genetics identifies variations in deoxyribonucleic acid (DNA) sequence that are associated with phenotypic variance. Basic descriptions of quantitative genetics and molecular genetics in relation to behavioral research are available elsewhere (Plomin, DeFries, et al., 2001).

Genetics
Intelligence was one of the first human traits to be the target of genetic research even before psychology emerged as a scientific field. A year before the publication of Mendels seminal article on the laws of heredity, Galton (1865) published a two-article series on high intelligence and other abilities, which he later expanded into the first book on heredity and intelligence (Galton, 1869). Galton (1883) provoked a needless battle that raged through the 20th century by arguing that there is no escape from the conclusion that nature prevails enormously over nurture (p. 241), especially because his research on family resemblance could not by itself disentangle genetic and environmental influences. Although Galton (1876) suggested the twin design and the adoption design, the first twin and adoption studies were not carried out until the 1920s each of these investigated intelligence (Burks, 1928; Freeman, Holzinger, & Mitchell, 1928; Merriman, 1924; Theis, 1924). The first animal model research on learning and problem solving
112

Robert Plomin and Frank M. Spinath, Social, Genetic and Developmental Psychiatry Research Centre, Institute of Psychiatry, Kings College London, London, England. Frank M. Spinath is now at the Department of Psychology, University of Bielefeld, Bielefeld, Germany. Correspondence concerning this article should be addressed to Robert Plomin, Social, Genetic and Developmental Psychiatry Research Centre, Box No. P080, Institute of Psychiatry, Kings College London, 111 Denmark Hill, London SE5 8AF, England. E-mail: [email protected]

SPECIAL SECTION: 100 YEARS AFTER SPEARMAN (1904)

113

was also relevant to individual differences in intelligence, most notably the successful selection study of maze-bright and mazedull rats bred initially by Tolman in 1924 and continued by Tryon (described by McClearn, 1963). In the 1950s and 1960s, studies of inbred strains of mice showed the important contribution of genetics to individual differences for most aspects of learning. In 1963, a review in Science of genetic research on intelligence was influential in showing the convergence of evidence from family, twin, and adoption studies pointing to genetic influence (Erlenmeyer-Kimling & Jarvik, 1963). During the 1960s, environmentalism was beginning to wane in psychology and the stage was set for increased acceptance of genetic influence on intelligence. Then, in 1969, a Harvard Educational Review monograph (Jensen, 1969) almost brought the field to a halt because it suggested that ethnic differences might involve genetic differences. Exactly 25 years later, this issue was resurrected in The Bell Curve (Herrnstein & Murray, 1994) and caused a similar uproar.

The storm raised by Jensens (1969) monograph led to intense criticism of all behavioral genetic research, but especially research in the area of intelligence (e.g., Kamin, 1974). These criticisms had the positive effect of generating about a dozen bigger and better behavioral genetic studies that produced much more data on the genetics of intelligence than had been obtained in the previous 50 years. Intelligence is the target of more genetic research than any other domain in science, with the exception of self-report personality questionnaires. Some of the new data and all of the old data were summarized in another influential Science article (Bouchard & McGue, 1981) that began to turn the tide in psychology toward acceptance of genetic influence on intelligence (Neisser et al., 1996; Snyderman & Rothman, 1987, 1988). Figure 1 summarizes the review and updates it (Plomin, DeFries, et al., 2001). For example, in studies of more than 10,000 monozygotic (MZ) and dizygotic (DZ) pairs of twins, the average MZ correlation is .86, which is near the testretest reliability of the measures, in contrast

Figure 1. Average IQ correlations for family, adoption, and twin designs. On the basis of reviews by Bouchard and McGue (1981) as amended by Loehlin (1989). Data for old data for monozygotic (MZ) twins exclude the suspect data of Burt (1966); the new data include Bouchard, Lykken, McGue, Segal, and Tellegen (1990) and Pedersen, McClearn, Plomin, and Nesselroade (1992). DZ dizygotic, P-O parent offspring; Sib sibling. From Behavioral Genetics (4th ed., Figure 9.7, p. 168), by R. Plomin, J. C. DeFries, G. E. McClearn, and P. McGuffin (Eds.), 2001, New York: Worth. Copyright 2001 by W. H. Freeman/Worth. Adapted with permission.

114

PLOMIN AND SPINATH

to the DZ correlation of .60. Adoption data, including adoptedapart parents and offspring and adopted-apart siblings as well as MZ twins adopted apart, also point to substantial genetic influence. Model-fitting analyses that simultaneously analyze all of the family, adoption, and twin data summarized in Figure 1 yield heritability estimates of about 50% (Chipuer, Rovine, & Plomin, 1990; Loehlin, 1989). In other words, about half of the total variance (which includes error of measurement) can be attributed to DNA differences between individuals. Heritability is of course higher if corrections are made for error of measurement as in analyses of latent variables free of measurement error. Even an attempt to explain as much of the variance of g as possible in terms of prenatal effects nonetheless yielded a heritability estimate of 48% (Devlin, Daniels, & Roeder, 1997; McGue, 1997). Although most of this research was conducted in the United States and Western European countries, similar estimates of heritability have been found in countries such as Moscow, the former East Germany, Japan, and rural and urban India (Plomin, DeFries, et al., 2001). The convergence of evidence on the conclusion that individual differences in intelligence are substantially heritable led to a decline in the 1990s of genetic research on intelligence that merely aimed to investigate the heritability of intelligence. Instead, genetic designs were used to go beyond estimating heritability in order to ask questions about environmental influences, developmental change and continuity, and multivariate issues. Before discussing these three topics, it should be noted that assortative mating for intelligence is substantial. Correlations between spouses are only about .10 for other personality traits and about .20 for height and weight, but assortative mating for intelligence is about .40 (Plomin, DeFries, et al., 2001). The importance of assortative mating is that it increases genetic variance generation after generation and may thus contribute to the high heritability of intelligence. Twin studies that do not take assortative mating into account underestimate heritability because the genetic effects of assortative mating inflate the DZ correlation but not the MZ correlation (Plomin, DeFries, et al., 2001).

as socioeconomic status are likely to contribute. There has been one largely unsuccessful attempt to identify specific aspects of the home environment that are responsible for the shared environmental influence on intelligence in childhood (Chipuer & Plomin, 1992). However, as explained in the following section on development, although shared environment is important for intelligence in childhood, its importance declines to negligible levels after adolescence. In other words, shared environmental factors relevant to intelligence would be expected to show associations in childhood but not later in development. Moreover, this finding suggests that, even for intelligence, the salient environmental factors are nonshared after childhood. The second finding has been called the nature of nurture (Plomin & Bergeman, 1991). When used as outcome measures in genetic research, environmental measures consistently point to some genetic influence, suggesting that genetic factors influence the way we react and interact with the environment, a type of genotype-environment correlation (Kendler & Eaves, 1986). For example, a widely used measure of the home environment in research on cognitive development is the Home Observation for Measurement of the Environment (HOME; Bradley, Convyn, Burchinal, McAdoo, & Coll, 2001). In an adoption study comparing nonadoptive and adoptive siblings, genetic influences were estimated to account for about 40% of the variance of HOME scores (Braungart, Fulker, & Plomin, 1992). Moreover, multivariate genetic analysis (described below) indicated that about half of the phenotypic correlation between the HOME and childrens intelligence is mediated genetically. This research suggests that we create our experiences in part for genetic reasons and supports a current shift from thinking about passive models of how the environment affects individuals toward models that recognize the active role we play in selecting, modifying, and creating our own environments (Plomin, 1994). In quantitative genetics, this topic is referred to as active genotype-environment correlation (Plomin, DeFries, et al., 2001).

Development Environment
Concerning the environment, genetic research provides the best available evidence for the importance of environmental influences on intelligence: If heritability is 50%, that means that environmental factors account for the rest of the reliable variance. Two of the most important findings from genetic research are about nurture rather than nature. First, nearly all personality traits show that, contrary to theories of socialization from Freud onwards, environmental influences operate to make siblings growing up in the same family as different from one another as children growing up in different families (Harris, 1998; Plomin & Daniels, 1987). However, intelligence is the exception to this rule of nonshared environmental influence (Plomin, 1988). Direct estimates of the importance of shared environmental influence come from correlations of .19 for adoptive parents and their adopted children and .32 for adoptive siblings (see Figure 1). Because adoptive siblings are unrelated genetically, what makes them similar is shared rearing, suggesting that about a third of the total variance can be explained by shared environmental influences. The factors responsible for this shared environmental influence have not been pinned down, although general family background variables such Two types of developmental questions have been addressed in genetic research. The first question is: Does heritability change during development? Because it is so reasonable to assume that genetic differences become less important as experiences accumulate during the course of life, one of the most interesting findings about intelligence is that the opposite is closer to the truth. Research during the past decade has shown that the heritability of g increases during development. Figure 2 summarizes twin results by age (McGue, Bouchard, Iacono, & Lykken, 1993), showing that the difference between MZ and DZ twin correlations increases slightly from early to middle childhood and then increases dramatically in adulthood. Because relatively few twin studies of intelligence have included adults, summaries of intelligence data (see Figure 1) showing heritability estimates of about 50% rest primarily on data from childhood. Heritability in adulthood is higher, perhaps as high as 80%, although there is some evidence that heritability late in life might be lower (Finkel, Pedersen, McGue, & McClearn, 1995). A finer grained analysis of twin results indicates that heritability is lower in infancy (about 20%) than in middle childhood (about 40%; Plomin, 1986). The modest heritability of intelligence in early childhood was confirmed in a

SPECIAL SECTION: 100 YEARS AFTER SPEARMAN (1904)

115

Figure 2. The difference between monozygotic (MZ) and dizygotic (DZ) twin correlations for intelligence increases during development suggesting increasing genetic influence. Adapted from Behavioral Genetics of Cognitive Ability: A Life-Span Perspective (Figure 1, p. 63, by M. McGue, T. J. Bouchard Jr., W. G. Iacono, & D. T. Lykken, in Nature, Nurture, and Psychology, by R. Plomin and G. E. McClearn (Eds.), 1993, Washington, DC: American Psychological Association. Copyright 1993 by the American Psychological Association.

recent study of nearly 7,000 twin pairs assessed longitudinally using parent-administered tests, which yielded heritability estimates of .27, .30, and .25 at 2, 3, and 4 years, respectively (Spinath, Ronald, Harlaar, Price, & Plomin, 2003). Results from a 16-year longitudinal adoption study support this view of increasing heritability (Plomin, Fulker, Corley, & DeFries, 1997). As shown in Figure 3, parent offspring correlations for parents (called control parents) who share both genes and environment with their offspring increase during childhood and adolescence, as has been found in other family studies, and this is not due to increased reliability of assessing infant intelligence (Bayley, 1969). The adoption design directly assesses the genetic contribution to this parent offspring resemblance by studying biological (birth) parents and their adopted-away offspring. Correlations between biological parents and their adopted-away offspring are similar to the correlations for control parents and offspring. In contrast, these adopted children show no resemblance to the parents who adopted them. Why does heritability of g increase during the life span? It is possible that completely new genes come to affect g as more sophisticated cognitive processes come on line during development. However, another hypothesis is that relatively small genetic effects early in life snowball during development, creating larger and larger phenotypic effects as individuals select or create environments that foster their genetic propensities (Plomin & DeFries,

1985). This hypothesis relates to the notion of active genotypeenvironment correlation, which was mentioned earlier. Another developmental finding of great importance concerns shared environmental influence. As noted earlier, intelligence, unlike other personality traits, shows shared environmental influence. The twin data summarized in Figure 2 suggest that shared environment effects are negligible in adulthood. Data for adoptive siblings summarized in Figure 1, which provides a direct test of shared environment, indicate substantial shared environmental influence. However, the studies of adoptive siblings summarized in Figure 1 assessed the adoptive siblings when they were children. Recent studies of adoptive siblings assessed after adolescence show an average correlation of zero (McGue et al., 1993). These results indicate that although shared environment affects intelligence in childhood, in the long run environmental influences on intelligence are nonshared. A second type of developmental question involves analyses of change and continuity using longitudinal data. Data of this type are analyzed using multivariate genetic analysis (described below), but a simple way to think about genetic contributions to developmental change is to ask whether changes in intelligence scores from age to age show genetic influence. Longitudinal research indicates that genetic factors account in part for such changes, especially in childhood (Fulker, Cherny, & Cardon, 1993) and perhaps even in adulthood (Loehlin, Horn, & Willerman, 1989), although most

116

PLOMIN AND SPINATH

Figure 3. Parent offspring correlations between parent and offspring intelligence scores for biological mothers and adoptive and control parents and their children at 3, 4, 7, 9, 10, 12, 14, and 16 years. Adoptive and control parent offspring correlations are weighted averages for mothers and fathers in order to simplify the presentation. The sample sizes range from 159 to 195 for biological mothers, 153 to 194 for adoptive parents, and 136 to 216 for control parents. From Nature, Nurture and Cognitive Development from 1 to 16 years: A ParentOffspring Adoption Study, by R. Plomin, D. W. Fulker, R. Corley, and J. C. DeFries, 1997, Psychological Science, 8, Figure 1, p. 443. Copyright 1997 by Blackwell Publishers. Adapted with permission.

genetic effects on intelligence contribute to continuity rather than change even late in life (Plomin, Pedersen, Lichtenstein, & McClearn, 1994). Unlike genetic effects, longitudinal genetic analysis suggests that shared environmental effects contribute only to continuity in childhood. In other words, some relatively constant factors such as the familys socioeconomic status might account for the developmental continuity of shared environmental influence on intelligence.

Multivariate Analysis
Specific cognitive abilities such as verbal ability, spatial ability, and memory show substantial genetic influence, although less than for general intelligence (Plomin & DeFries, 1998). To what extent do different sets of genes affect different abilities? We know that diverse cognitive tests correlate moderatelythis is the basis for Spearmans g. A meta-analysis of all cognitive studies yielded an average correlation of about .30 (Carroll, 1993), although studies using less restricted samples and more reliable measures yield higher intercorrelations (Jensen, 1998). A technique called multivariate genetic analysis can be used to examine genetic and environmental contributions to the phenotypic covariance among

specific cognitive abilities (Plomin, DeFries, et al., 2001). Multivariate genetic analysis yields a statistic called genetic correlation, which is an index of the extent to which genetic effects on one trait correlate with genetic effects on another trait independent of the heritability of the two traits. That is, two traits could be highly heritable but the genetic correlation between them could be zero. Conversely, two traits could be only modestly heritable but the genetic correlation between them could be 1.0, indicating that even though genetic effects are not strong (because heritability is modest) the same genetic effects are involved in both traits. In the case of specific cognitive abilities that are moderately heritable, multivariate genetic analyses have consistently found that genetic correlations are very high close to 1.0 (Petrill, 1997). That is, although Spearmans g is a phenotypic construct, g is even stronger genetically. These multivariate genetic results predict that when genes are found that are associated with one cognitive ability, such as spatial ability, they will also be associated just as strongly with other cognitive abilities, such as verbal ability or memory. Conversely, attempts to find genes for specific cognitive abilities independent of general cognitive ability are unlikely to succeed because what is in common among cognitive abilities is

SPECIAL SECTION: 100 YEARS AFTER SPEARMAN (1904)

117

largely genetic and what is independent is largely environmental. Identifying genes associated with cognitive abilities will test the hypothesis that the same genes affect diverse cognitive abilities. This finding of substantial genetic overlap among cognitive abilities also has important implications for understanding the brain mechanisms that mediate genetic effects on intelligence. In contrast to the prevalent modular view of cognitive neuroscience that assumes that cognitive processes are specific and independent, these results suggest that genetic effects are general (Plomin & Spinath, 2002). Recent multivariate genetic research on so-called elementary cognitive processes thought to underlie general intelligence suggests that genetic correlations are just as strong among these elementary cognitive processes. In other words, the genetic version of Spearmans g also emerges at the level of elementary cognitive processes. It might also exist in the brain. For example, recent twin studies using magnetic resonance imaging to assess brain volume find that brain volume is highly heritable, substantially intercorrelated across brain regions, and moderately correlated with intelligence (Pennington et al., 2000; Thompson et al., 2001). Although the reductionistic model of brain 3 cognition 3 behavior is deeply embedded in our thinking, an agnostic model in which brain, cognitive, and psychometric measures are considered merely as correlates rather than causes of the genetic version of Spearmans g is all that is warranted from the data so far (Plomin & Spinath, 2002). A second issue concerns the relationship between the normal and abnormal. For example, to what extent is mild mental retar-

dation (MMR) genetically distinct from the rest of the distribution of intelligence? Surprisingly, no twin or adoption studies of MMR have been reported until recently (see the next paragraph). More than 200 rare single-gene disorders include mental retardation, often severe retardation, as a symptom (Zechner et al., 2001), and many chromosomal causes of mental retardation are also known (Plomin, DeFries, et al., 2001), including microdeletions of bits of chromosomes (Baker et al., 2002; Knight et al., 1999). In general, many of the single-gene mutations tend to be spontaneous in the affected individual as are most of the chromosomal anomalies. That is, these DNA causes of severe mental retardation are not usually inherited. Although no twin studies of severe mental retardation have been reported, an interesting sibling study shows no familial resemblance. In a study of over 17,000 children, 0.5% were moderately to severely retarded (Nichols, 1984). As shown in Figure 4 (dotted line), siblings of these retarded children were not retarded. The siblings average IQ was 103, with a range of 85 to 125. In other words, moderate to severe mental retardation showed no familial resemblance, a finding implying that mental retardation is not heritable. In contrast, siblings of mildly retarded children (1.2% of the sample) tend to have lower than average IQ scores (see Figure 4, solid line). The average IQ for these siblings of mildly retarded children was only 85. Similar findingsthat MMR is familial but moderate and severe retardation are not familialalso emerged from the largest family study of MMR, which considered 80,000 relatives of 289 mentally retarded individuals (Reed & Reed, 1965).

Figure 4. Siblings of severely retarded children tend to have normal IQs, whereas siblings of mildly retarded children tend to have lower than normal IQs. These results suggest that mild retardation is familial and perhaps heritable but severe retardation is not. From Behavioral Genetics (4th ed., p. 162) by R. Plomin, J. C. DeFries, G. E. McClearn, and P. McGuffin, 2001, New York: Worth Publishers. Copyright 2001 by W. H. Freeman/ Worth. Reprinted with permission. From Familial Mental Retardation, by P. L. Nichols, 1984, Behavior Genetics, 14, p. 167. Copyright 1984 by Kluwer Academic. Adapted with permission.

118

PLOMIN AND SPINATH

The vast majority of quantitative genetic research on disorders focuses on the dichotomous diagnosis itself rather than on a quantitative trait. Using a diagnostic or quantitative cutoff, most studies assess concordancesthe risk that a family member of a proband will also have the disorder. For example, the first large twin study of MMR selected children from the lowest 5% of the distribution from a representative sample of 3,886 twins in samesex and opposite-sex pairs on the basis of an aggregate intelligence score obtained at 2, 3, and 4 years of age (Spinath, Harlaar, Ronald, & Plomin, in press). Twin concordances were 74% for MZ twins, 45% for DZ same-sex (DZS) twins, and 36% for DZ opposite-sex (DZO) twins, suggesting substantial genetic influence for the lowest end of the distribution of intelligence. However, unlike twin similarity coefficients such as MZ and DZ intraclass correlations, concordances cannot be used to estimate heritability simply by doubling the difference in MZ and DZ concordances. Instead, concordances are typically converted into liability (tetrachoric) correlations on the assumption that a continuum of genetic risk or liability underlies the dichotomous diagnosis and that the disorder is seen only when a certain threshold of liability is exceeded (Falconer, 1965; Neale, Boker, Xie, & Maes, 1999). Rather than assessing a dichotomy and assuming a continuum with a threshold, however, it is advantageous to use the continuum of IQ scores directly. An approach that uses quantitative scores to assess family resemblance for a dichotomous diagnosis has been systematized as an analytic strategy that is called DF extremes analysis (DeFries & Fulker, 1985, 1988). DF extremes analysis assesses the extent to which the mean quantitative trait score of cotwins of MMR probands differs from the population mean on the quantitative trait. Group familiality is indicated to the extent that the cotwin mean is closer to the proband mean than to the population mean. This familial resemblance is called group familiality because it refers to the familial origins of the average difference between the probands and the population, not to individual differences among the probands. Genetic influence on the mean quantitative trait score difference between the probands and the population is indicated to the extent that the average quantitative trait score of MZ cotwins is more similar to the proband mean than is the DZ cotwin mean. This estimate of heritability is called group heritability because it indicates the extent to which the mean quantitative trait score difference between probands and the population can be explained by genetic factors. Initially conceptualized as a regression method that was limited to the analysis of same-sex twin pairs, the basic DF model has recently been reframed in model-fitting terms allowing DZO to be incorporated in a sex-limitation model that tests for sex differences in genetic and environmental parameters (Purcell & Sham, 2003). In the study of MMR in young children mentioned earlier, the mean quantitative trait score of DZS and DZO cotwins as compared with MZ cotwins regressed farther toward the population mean, suggesting genetic influence on the mean quantitative trait score difference between the MMR probands and the unselected population. DF extremes analysis yielded a group heritability (h2g) estimate of .49 (CI .29, .69). In other words, about half of the difference between the MMR probands and the population can be attributed to genes. This group heritability estimate of .49 is significantly greater than the heritability of individual differences in the sample from which the MMR probands were selected, which is a typically

modest heritability estimate for early childhood (h2 .24, CI .21, .27), suggesting that genetic factors have a stronger effect at the low end of the distribution. This finding does not necessarily mean that different genes affect MMR and the rest of the distribution of intelligence. For example, it is possible that the environment has less impact on individuals at high genetic risk. The extent to which the same genes affect MMR and the rest of the distribution will not be known definitively until specific genes are found for MMR or intelligence. However, finding group heritability implies that there is a genetic relationship between MMR and individual differences in the quantitative trait across the normal range. For this reason, it seems likely that MMR is at the lower end of the distribution of genetic and environmental factors that are responsible for individual differences in intelligence, despite the higher heritability for MMR (Plomin, 1999). Research in this same vein on high intelligence suggests that group heritability is similar to the heritability of individual differences in intelligence (Plomin & Price, 2003; Ronald, Spinath, & Plomin, 2002).

Genes
The 20th century began with the rediscovery of Mendels laws of heredity. The word gene was first coined in 1903. Fifty years later the double helix structure of DNA was discovered. The genetic code was cracked in 1966 the four-letter alphabet (G, A, T, C) of DNA is read as three-letter words that code for the 20 amino acids that are the building blocks of proteins. The crowning glory of the century and the beginning of the new millennium is the Human Genome Project, which has provided a working draft of the sequence of the 3 billion letters of DNA in the human genome, nucleotide bases that are the steps in the spiral staircase of DNA (International Human Genome Sequencing Consortium, 2001; Venter et al., 2001). Progress is also being made toward identifying all of the genes from the genome sequence. In the traditional sense of the central dogma of DNA, a gene is DNA that is transcribed into ribonucleic acid (RNA), and then translated into amino acid sequences. Less than 2% of the more than 3 billion bases of DNA in the human genome are of this sort. It is not yet known how many such genes there are in the human genome. It used to be said that there are 100,000 genes, but the 2001 working draft of the human genome suggested far fewer, perhaps as few as 30,000, although estimates of the number of genes have been rising again subsequently. Moreover, some of the other 98% of DNA may be important. One example is DNA that is transcribed into RNA but not translated. For nearly all genes, a complicated process called splicing occurs between transcription and translation. All of the DNA within a gene is transcribed into RNA, but segments of RNA (called introns) are deleted and remain in the nucleus whereas the other segments (called exons) are spliced back together and exit the nucleus where they are translated into amino acid sequences. Although introns have been thought to be genetic junk that have hitched a ride evolutionarily, it is now known that in some cases introns regulate the transcription of other genes (Mattick, 2001). Exons are conserved evolutionarilymost of our exons are highly similar to DNA sequences in primates, mammals, and even invertebrates. This implies that the sheer number of such genes is not responsible for the greater complexity of the human species. More subtle variations in DNA rather than the number of genes may be

SPECIAL SECTION: 100 YEARS AFTER SPEARMAN (1904)

119

responsible for differences between mice and men (Brett, Pospisil, Valcarcel, Reich, & Bork, 2002). If subtle DNA differences are responsible for the differences between mice and men, even more subtle differences are likely to be responsible for individual differences within the human species. Although many single-gene disorders involve mutations in exons, introns might be sources of quantitative trait loci (QTLs) that have more subtle effects on gene regulation. Another example indicating that there is still much to learn about genes is the recent discovery of RNA genes called microRNA (Eddy, 2001). Rather than just encoding proteins, some very short noncoding RNA sequences produce functional RNA molecules that seem to be especially important in regulating gene expression. For behavioral genetics, the most important next step is the identification of the DNA sequences that make us different from each other. There is no single human genome sequencewe each have a unique genome. Most of the DNA letters in the four-letter alphabet of DNA are the same for all human genomesand many of these are the same for other primates and other mammals and even insects. Nevertheless, about one in every thousand DNA letters differs among people with at least 1% frequency, which means that there are about 3 million DNA variations in total, enough to make us each differ for almost every gene. Most of these DNA differences involve a substitution of a single base pair, called single nucleotide polymorphisms (SNPs, pronounced snips). These DNA differences are responsible for the widespread heritability of behavioral disorders and dimensions. That is, when we say that a trait is heritable, we mean that variations in DNA exist that cause differences in behavior. Particularly useful are SNPs in coding regions (cSNPs) that result in differences in the amino acid sequences coded by DNA and other SNPs that are potentially functional such as SNPs in DNA control regions that regulate the transcription of genes. The major beneficiary of these advances in molecular genetics will be research on complex traits such as intelligence that are influenced by multiple genes as well as multiple environmental influences. One of the most exciting directions for genetic research on intelligence is to harness the power of the Human Genome Project to begin to identify specific genes responsible for the heritability of intelligence. It should be noted that DNA variation has a unique causal status in explaining behavior. When behavior is correlated with anything else, the old adage applies that correlation does not imply causation. For example, although aspects of the family environment correlate with childrens intelligence, this correlation is not necessarily causal. As mentioned earlier, behavioral genetic research has shown that family environment in part reflects genetic influences on childrens intelligence. When it comes to interpreting correlations between biology and behavior, such correlations are often mistakenly interpreted as if biology causes behavior. For example, correlations between neurotransmitter physiology and behavior or between neuroimaging indices of brain activation and behavior are often interpreted as if brain differences cause behavioral differences. However, these correlations do not necessarily imply causation because behavioral differences can cause brain differences. In contrast, in the case of correlations between DNA variants and behavior, the behavior of individuals does not change their genome. Expression of genes can be altered but the DNA sequence itself does not change (except in the evolutionary sense of natural selection). For this reason, correlations between DNA

differences and behavioral differences can be interpreted causally: DNA differences can cause the behavioral differences but not the other way around.

Linkage
The first generation of DNA research, which began in the 1980s, focused on the thousands of rare single-gene disorders, such as Huntingtons disease, in which a single gene is necessary and sufficient for the emergence of the disorder. The heritability of such single-gene disorders is 100%, which makes their localization on a chromosome and then the ultimate pinpointing of a particular DNA sequence relatively straightforward. The standard approach, first used successfully to localize the gene for Huntingtons disease to the tip of chromosome 4 in 1983 (Gusella et al., 1983), was to look for linkage in large family pedigrees between the disease and one of a few hundred DNA markers evenly spread throughout the chromosomes. Linkage is a violation of Mendels second law of independent assortment that posits that two traits will be inherited independently. Mendel did not know that genes are on chromosomes. If two genesfor example, a gene for a disorder and a DNA markerare close together on a chromosome, they may be inherited as a package within families rather than independently as predicted by Mendels second law. For example, the linkage of Huntingtons disease with DNA markers was found in a single five-generation family of hundreds of individuals when a particular form (called allele) of a DNA marker on chromosome 4 was only found in family members who had Huntingtons disease. Similar linkage studies have identified the chromosomal location of hundreds of single-gene disorders and the precise DNA fault has been found for many of these disorders. Linkage studies of this type were also undertaken for psychiatric disorders even though there was no suggestion that such complex disorders are inherited as single-gene disorders. Early successes were claimed for bipolar depression (Egeland et al., 1987) and for schizophrenia (Sherrington et al., 1988), but neither claim was replicated. It is now clear that this traditional linkage approach can only detect a linkage if the gene has a very large effect on the disorder, a situation limited to relatively rare disorders such as Huntingtons disease, which has a frequency of about 1 in 20,000 individuals. Common disorders including cognitive problems such as learning disabilities in childhood and dementia late in life seldom show any sign of single-gene effects and appear to be caused by multiple genes as well as by multiple environmental factors. Indeed, quantitative genetic research mentioned earlier suggests that such common disorders are usually the quantitative extreme of the same genes responsible for variation throughout the distribution (Plomin, Owen, & McGuffin, 1994). Genes in such multiple-gene systems are called QTLs because they are likely to result in dimensions (quantitative continua) rather than in disorders (qualitative dichotomies). In other words, in terms of the genetic etiology of common disorders, there may be no disorders, just dimensions. The QTL perspective is the molecular genetic extension of quantitative genetics in which genetic variation is viewed as normal and is distributed quantitatively. The goal of QTL research is not to find the gene for a complex trait, but rather the multiple genes that make contributions of varying effect sizes to the variance of the trait. Perhaps one gene will be found that accounts for 5% of the variance, 5 other genes

120

PLOMIN AND SPINATH

might each account for 2% of the variance, and 10 other genes might each account for 1% of the variance. If the effects of these QTLs are independent, the QTLs would in total account for 25% of the traits variance. All of the genes that contribute to the heritability of a complex trait are unlikely to be identified because some of their effects may be too small to detect. The problem is that we do not know the distribution of effect sizes of QTLs for any complex trait in plant, animal, or human species. Not long ago, a 10% effect size was thought to be small, at least from the single-gene perspective in which the effect size was essentially 100%. However, for behavioral disorders and dimensions, a 10% effect size may turn out to be a very large effect. If effect sizes are as small as 1%, this would explain the slow progress to date in identifying genes associated with behavior because research so far has been woefully underpowered to detect and replicate QTLs of such small effect size (Cardon & Bell, 2001). Recent research has been more successful in finding QTLs for complex traits because designs have been used that can detect genes of much smaller effect size. The problem with the traditional large-pedigree linkage method in relation to intelligence and other personality traits is that there is no dichotomous disorder that can be used to chart coinheritance with DNA markers within families. There have been no traditional linkage studies of intelligence or other quantitative traits, although, as mentioned earlier, linkage has been successful in leading to the identification of more than 200 rare single-gene syndromes for which mental retardation is a symptom (Zechner et al., 2001). Linkage has been extended to consider QTLs by using many small families (usually siblings) rather than a few large families. These QTL linkage methods can be used to study the extremes of a quantitative trait or a diagnosed disorder and are able to detect genes that account for about 10% of the variance of the quantitative trait or the assumed liability or susceptibility to the disorder. The essence of the most popular method, called sib-pair QTL linkage analysis, is to ask whether sharing alleles for a particular DNA marker makes siblings more similar phenotypically. Siblings can share zero, one, or two of the alleles that they inherit from their parents. Thus, in relation to a particular DNA marker, a pair of siblings can be like adoptive siblings sharing no alleles, like DZ twins sharing one allele, or like MZ twins sharing the same two alleles. An analysis similar to the twin analysis can be used to analyze the extent to which allele sharing affects sibling phenotypic resemblance. This method was used to identify the first QTL linkage, which was a linkage for reading disability (Cardon et al., 1994), a linkage that has been consistently replicated in several studies (Willcutt et al., 2003). Other QTL linkages for reading have also been reported (S. Fisher, 2003).

in length. In both studies, individuals with long DRD4 alleles had significantly higher novelty-seeking scores than did individuals with short alleles. The distributions of novelty-seeking scores for individuals with the short and the long DRD4 alleles show that the effect is small, accounting for about 4% of the variance in this sample. As would be expected for an association of small effect, many studies have failed to replicate the association, although at least a dozen studies have found it (Prolo & Licinio, 2002). Considering the power of the studies, the results are consistent with a QTL that accounts for about 1% of the variance (Plomin & Caspi, 1998). The vast majority of association studies involves case-control comparisons for diagnosed disorders. For example, DRD4 also shows an association with hyperactivity in the expected directionlong alleles are associated with greater risk for hyperactivity (Thapar, 2003). Of 15 published studies, 11 have found evidence of association comparing cases and controls, and a meta-analysis indicates a significant effect with an odds ratio of about 2 (Faraone, Doyle, Mick, & Biederman, 2001). The first such association with a disease was identified for late-onset Alzheimers disease in 1993 (Corder et al., 1993) and has been replicated in scores of studies (Williams, 2003). The gene is apolipoprotein E (APOE), which codes for a serum lipoprotein involved in cholesterol metabolism. One of the APOE alleles (APOE-4) has a frequency of about 40% in individuals with late-onset Alzheimers disease and about 15% in controls. In a meta-analysis of 40 studies involving 15,000 individuals, elevated frequencies of APOE-4 was found for Alzheimers patients in each study, although the association was stronger among Caucasians and Japanese and weaker in African Americans (Farrer et al., 1997). APOE has a large effect for a QTL, but it is a QTL in that the APOE-4 allele is by no means necessary or sufficient for the development of the disorderit is a risk factor that increases susceptibility to the disorder. At least one third of individuals with Alzheimers disease lack the allele, and up to half of individuals who have a double dose of this allele survive to age 80 without developing the disease (Williams, 2003). It sounds contradictory to refer to a QTL association with a dichotomous disorder such as Alzheimers disease because diagnosed disorders are not quantitative traits. However, if several genes affect the disorder, which is implied if a particular gene has a small effect, the genes will produce a continuum of susceptibility to the disorder. This implies that the disorder is actually the extreme of a dimension, as discussed earlier. Much ongoing QTL research on intelligence is coming from the intense research effort on dementia, which usually assesses intelligence prior to the decline of dementia.

Association
A second method, called allelic association, can detect QTLs that account for much smaller amounts of variance than linkage (Risch, 2000; Risch & Merikangas, 1996). Association is also simpler than linkage: Association is the correlation between a particular allele and trait in the population. For example, one of the first associations reported for personality was an association between the neuroreceptor gene, dopamine D4 receptor (DRD4), and novelty seeking (Benjamin et al., 1996; Ebstein et al., 1996). The DNA marker in the DRD4 gene has two types of alleles that vary

Association Studies of Intelligence

Intelligence is a reasonable target for QTL research for three reasons. First, it is substantially heritable. More interestingly, multivariate genetic research reviewed above indicates that intelligence is the level at which genes affect cognitive abilities, much more so than at the level of specific cognitive abilities. The third reason is that QTLs have been found for the cognitive disorders of reading disability and dementia and more than 200 rare single gene disorders have been isolated that include mental retardation among their symptoms.

SPECIAL SECTION: 100 YEARS AFTER SPEARMAN (1904)

121

One project called the IQ QTL Project has systematically attempted to identify QTLs associated with intelligence (Plomin, Hill, et al., 2001). Although no solid QTL associations have yet emerged, the IQ QTL Project is described in this section because the project raises many current issues relevant to the pursuit of specific genes associated with any complex traits, not just intelligence. For example, the IQ QTL Project is the first molecular genetic study to focus on ability rather than disability. This focus highlights the point that genetic variation occurs throughout the distribution genetic variation does not just consist of rare mutations that cause severe disorders. Rather than using the entire distribution as in QTL studies of other personality traits (Benjamin et al., 2002), the IQ QTL Project selected very high-functioning individuals in order to increase power to detect QTLs of small effect size. Its goal is not to find genes for genius but rather to use very high-functioning individuals in order to identify QTLs that operate throughout the entire distribution, including the low (MMR) end of the ability distribution. This approach is based on the simple hypothesis that, although any one of many genes can disrupt normal development, very high functioning requires most of the positive alleles and few of the negative alleles. This is just a hypothesis, but one that can be tested when QTLs are found because it predicts that QTLs found for high ability will have a similar effect throughout the rest of the distribution including the low end of the distribution. The IQ QTL Project currently includes an original sample of 101 cases with mean IQ of 136 and 101 controls with mean IQ of 100. The high group for the original sample, which is more than two standard deviations above the population mean, represents the equivalent of the top 2% of an unselected sample of 5,000 individuals. Because greater power is needed to replicate results, a replication sample included 96 individuals of some of the brightest adolescents in the United States with estimated IQs greater than 160 (equivalent to the top .00003 of an unselected sample of 3 million) as well as another sample of 100 controls with mean IQ of 100. For QTLs with 5%, 2.5%, and 1% effect sizes, the original sample provides 100%, 93%, and 56% power, respectively; the replication sample provides even greater power because the sample is more extreme: 100%, 100%, and 99%, respectively. However, these power estimates assume that the marker is very close to the QTLpower drops off rapidly as the distance between the marker and the QTL increases. The first phase of the project, which had much smaller sample sizes, consisted of genotyping 100 DNA markers in or near genes involved in brain functioning (Plomin et al., 1995). For example, the association between APOE-4 and dementia makes this gene a reasonable candidate for association with intelligence. There was a suggestion of an association involving the APOE gene in the expected direction, with the APOE-4 allelewhich is in higher frequency in individuals with dementiashowing a lower frequency in the high-intelligence group. However, a follow-up analysis that included a sample twice as large as the original found little evidence for association (Turic, Fisher, Plomin, & Owen, 2001). The earlier survey of 100 markers also included two markers for the catechol-o-methyltransferase gene (COMT) that did not suggest associations (Plomin et al., 1995). The COMT gene has been reported recently to correlate with working memory (Egan et al., 2001), which is highly correlated with intelligence (Deary, 2001).

One problem with such a candidate gene approach is that many of the thousands of genes expressed in the brain could be considered as candidate genes for intelligence. Allelic association can be made more systematic by using a dense map of markers. The IQ QTL Project took a first step in this direction by genotyping 47 DNA markers on the long arm of chromosome 6 (Chorney et al., 1998). A replicated association for a marker was found, which happened to be in the gene for insulin-like growth factor-2 receptor (IGF2R), which has been shown to be especially active in brain regions most involved in learning and memory (Wickelgren, 1998). We replicated this result using larger samples and using a different polymorphism in IGF2R (Hill, Chorney, et al., 1999). However, this QTL association did not hold up when another large independent sample was analyzed (Hill, Chorney, & Plomin, 2002). The problem with using a dense map of markers for a genome scan for QTLs of small effect is the amount of genotyping required. The number of markers needed for a complete genome scan is a matter of some uncertainty (Abecasis et al., 2001; Kruglyak, 1999; Reich et al., 2001), but it seems likely that at least 100,000 markers would be needed. Furthermore, these thousands of markers would need to be genotyped for large samples in order to detect QTLs of small effect size. For example, in order to detect a QTL of 1% heritability in an unselected sample with 80% power ( p .05), 800 individuals would need to be genotyped (Cohen, 1988). However, in order to protect against false positive results caused by genotyping so many markers, much lower alpha values and much larger samples are needed, resulting in the need for millions of genotypings when samples consist of thousands of individuals. To address these issues, the IQ QTL Project developed DNA pooling (J. Daniels, Holmans, Plomin, McGuffin, & Owen, 1998). DNA pooling greatly reduces the need for genotyping by pooling DNA from all individuals in a group and genotyping the pooled groups. In the IQ QTL Project, DNA is pooled for the 101 cases and for the 101 controls so that genotyping a single marker involves just two genotypings rather than 202 genotypings. In other words, for the cost of individually genotyping one marker for 200 individuals, DNA pooling makes it possible to genotype 100 markers for two groups of 100 individuals each. DNA pooling is a sensitive method for detecting the largest differences in allele frequencies between samples, as confirmed by individual genotyping (J. Daniels et al., 1998; Norton et al., 2002). As an example, Figure 5 shows DNA pooling results from the IQ QTL Project for a marker on chromosome 2. The markers used in the project are called simple-sequence repeat (SSR) markers because their alleles consist of short sequences of two, three, or four DNA bases that repeat a variable number of times. For example, the marker (D2S427) shown in Figure 5 includes a sequence of four DNA bases (GATA) that repeats typically 10, 11, 12, 13, 14, or 15 times. The function of such SSR markers, of which there are tens of thousands in the genome, is not known but the number of repeats is stably inherited and can be used as a DNA marker. Although each individual has two alleles, the genotyping results in Figure 5 show five alleles for all poolsthese are the bumps in Figure 5, which indicate the length of the DNA sequence and thus the number of repeats because the DNA was pooled across individuals in each group and thus represents all of their alleles. The heights of the bumps indicate the relative frequencies

122

PLOMIN AND SPINATH

Figure 5. DNA pooling results from the IQ QTL Project in the (A) original and (B) replication case-control samples for a tetranucleotide DNA marker (i.e., the polymorphism involves the number of repeats of a four base-pair motif) on chromosome 2 (D2S427) for the high-intelligence group (middle), control group (top), and their overlaid images (bottom). In all of the pools, five alleles for D2S427 are represented by the number of repeat units in the markerthe numbers above the allele image patterns (AIPs) represent the size (number of DNA base pairs) of the markers alleles, each of which differ by four base pairs. The relative frequencies of the alleles are represented by the height of each bump as indicated by the numbers below and to the right of the AIPs. (A sixth allele 262 base-pair units in length has a low frequency and can be seen only in the original high g pool.) The differences in the vertical scales of Panels A and B are due to differences in the amount of DNA that was amplified. The difference in the AIPs (AIP) for the two groups was calculated from the overlaid images by measuring the total area that was not shared by the two images irrespective of how many times the curves from the two pools crossed. This was then expressed as a fraction of the total shared and nonshared area (J. Daniels et al., 1998). The AIP simulated p values that test an overall difference in allele frequencies between the groups is .026 for the original sample and .003 for the replication sample. An allele-specific test of allele 2 (246 bp) comparing the high group and the control group yielded significant differences in the original sample but not in the replication sample. From A Genome-Wide Scan of 1842 DNA Markers for Allelic Associations With General Cognitive Ability: A Five-Stage Design Using DNA Pooling, by R. Plomin, L. Hill, I. Craig, P. McGuffin, S. Purcell, P. Sham, D. Lubinski, L. Thompson, P. J. Fisher, D. Turic, and M. J. Owen, 2001, Behavior Genetics, 31, p. 504. Copyright 2001 by Kluwer Academic. Adapted with permission.

SPECIAL SECTION: 100 YEARS AFTER SPEARMAN (1904)

123

of the alleles because the heights represent the number of copies of each allele in the pooled group. (For the original high g pool, a sixth low-frequency allele can be seen.) For this D2S427 DNA marker, the second allele (which is 246 base pairs in length, 4 base pairs longer than the first allele) indicates a lower frequency in the high-intelligence group than in the control group. This pattern was found in both the original sample (Panel A) and the replication sample (Panel B). When DNA pooling nominates a marker in this way, individual genotyping can be conducted to confirm the association. Individual genotyping for D2S427, shown in Figure 6, confirms the pattern of results suggested by DNA pooling although the association did not reach significance in the replication sample. Proof-of-principle articles for a systematic search of the genome using DNA pooling were published for chromosome 4 (P. J. Fisher et al., 1999) and chromosome 22 (Hill, Craig, et al., 1999), with samples only half the size of the current analyses. The first

Figure 6. Individual genotyping results for the marker D2S427, whose DNA pooling results are shown in Figure 5. Individual genotyping picked up a seventh rare allele, whereas DNA pooling detected only six alleles. An allele-specific test of allele 2 (246 bp) comparing the high group and the control group yielded significant differences in the original sample and the replication sample. Ori original; Rep replication; bp base pairs. From A Genome-Wide Scan of 1842 DNA Markers for Allelic Associations With General Cognitive Ability: A Five-Stage Design Using DNA Pooling, by R. Plomin, L. Hill, I. Craig, P. McGuffin, S. Purcell, P. Sham, D. Lubinski, L. Thompson, P. J. Fisher, D. Turic, and M. J. Owen, 2001, Behavior Genetics, 31, p. 505. Copyright 2001 by Kluwer Academic. Adapted with permission.

systematic genome scan for association for intelligence, using the complete samples described above, examined 1,842 markers (Plomin, Hill, et al., 2001). DNA pooling was used to screen the markers in the original and replication samples and then markers that met multiple conservative criteria for acceptance were individually genotyped in the two samples. DNA pooling of the original sample yielded 108 markers that yielded differences between cases and controls. Of these 108 markers, 6 met criteria using DNA pooling in the replication sample. These six markers were genotyped individually for the 202 subjects in the original sample and four yielded significant associations, which supports the validity of DNA pooling. The four markers that were significant in the original sample were then genotyped individually for the 196 subjects in the replication sample. Two of these four markers were also significant in the replication sample, markers on chromosome 4 (D4S2460) and chromosome 14 (D14S65). Although these two replicated QTL associations are noteworthy, when so many markers are genotyped, the possibility of false positive associations remains until the association is replicated in other studies. Moreover, a concern about case-control studies is that demographic differences, most notably ethnic differences, between the cases and controls might yield false positive results. Although all subjects in the IQ QTL Project are Caucasian, it is nonetheless possible that QTL associations could be due to some hidden ethnic stratification. For this reason, replication was sought in a third sample consisting of 196 parent child trios in which the offspring had estimated IQs greater than 160, which provides a withinfamily analysis called the Transmission Disequilibrium Test (TDT) that protects against population stratification as a possible source of QTL associations. When D4S2460 and D14S65 were genotyped individually for the 784 individuals in the 196 parent child trios, neither was significant. Although it is possible that the case-control results were caused by some hidden ethnic stratification within the Caucasian samples, another possibility is that the TDT lacks the power of the case-control design to detect QTLs of small effect. Nonetheless, the failure of the two candidate QTL associations to replicate in the TDT analysis led the IQ QTL Project to the conservative conclusion that this initial genome scan had not identified any clear QTL associations. Several markers that were nearly significant at all stages are being explored further in additional replication samples. The criteria for replication used in this study were extremely conservative. For example, no other QTL study has demanded replication in three samples using two different designs (casecontrol and parent offspring trios). However, a conservative approach is warranted given frequent failures to replicate QTL association results (Cardon & Bell, 2001). Although the multiplestage design with three extreme selected samples attempted to balance false positives and false negatives in an effort to detect QTLs of small effect size, the balance is in fact tilted very much in favor of avoiding false positives than false negatives for two reasons. The first reason is that the study only had power of about 50% to detect QTLs with 1% heritability, as mentioned earlier. For this reason these results should not be taken as an indictment of the QTL approach but rather they should serve as a warning of the exorbitant demands needed to obtain sufficient power to detect QTLs of small effect size.

124

PLOMIN AND SPINATH

A second reason why the balance was tilted against finding QTLs is that it is now known that perhaps a hundred thousand or even more DNA markers may be needed to exclude QTL association. As mentioned earlier, the problem for allelic association analysis is that power drops off rapidly when a marker is not very close to the QTL. For this reason, the IQ QTL Project is now focusing on potentially functional polymorphisms that may themselves be QTLs. However, rather than focusing on a few candidate genes or gene systems, we can look forward in the near future to a systematic search using all functional polymorphisms in coding sequences (cSNPs) and in regulatory regions. These tens of thousands of cSNPs can be genotyped using high-throughput techniques such as DNA pooling. In the meantime, the IQ QTL Project is using, as they become available, cSNPs that result in an amino acid substitution (called nonsynonymous cSNPs) as well as functional SNPs in regulatory regions. Although the problem remains that a QTL association with a functional SNP might in fact be due to another nearby SNP, it is a reasonable assumption that a functional SNP is the QTL, which greatly increases the power of QTL association. In addition to DNA pooling, other highthroughput techniques can genotype thousands of genes simultaneously, although these techniques are much more expensive than DNA pooling when used to genotype the large samples needed for QTL analysis (Craig & McClay, 2003). Other developments will also make genome-wide QTL searches more feasible. For example, haplotype mapping is a recent development that will greatly reduce the number of markers needed for a genome scan (Goldstein, 2001). Groups of contiguous markers across a chromosome are inherited as a package for hundreds of generations so that strategic selection of a few key markers can identify these chunks of chromosomes rather than genotyping all the other markers within a chunk. The road ahead will be much more difficult than generally assumed if, as we suspect, there are many QTLs associated with intelligence, which means that QTLs may account for less than 1% of the variance. Although the distribution of effect sizes is not known for intelligence or for any other complex trait, if QTL heritabilities are less than 1%, it will be difficult to detect them reliably. Nonetheless, the convergence of evidence for the strong heritability of intelligence from family, twin, and adoption studies indicates the existence of DNA polymorphisms associated with intelligence. The solution is that power will need to be increased in order to track down the QTLs responsible for the heritability of intelligence even if the QTL heritabilities are less than 1%. DNA pooling will be useful in this context because it costs no more to genotype 1,000 individuals than 100 individuals.

Three huge areas of functional genomic research have emerged: gene manipulation, gene expression profiling, and proteomics. Nearly all of this research has been conducted using animal models, primarily the mouse (Silver, 1995). Mouse models of personality traits such as activity and emotionality have been available for decades and, surprising as it may sound, mouse models of intelligence are also being developed (Plomin, 2001). These models use Spearmans g as a key. That is, g can be sought in the covariation among diverse measures of learning and memory and puzzle solving (Galsworthy, Paya-Cano, Monleon, & Plomin, 2002). Mouse models of intelligence are especially valuable for functional genomic research because, unlike the human species, both the genome and the environment can be controlled and manipulated. In addition, although the specificity of neuroimaging in humans is improving rapidly, much more precise analyses of brain function, such as single-cell recordings, can be conducted in mice.

Gene Manipulation: Targeted Mutations, Mutagenesis, and Antisense DNA

One way to study how a gene works is to knock it out by breeding mice for whom DNA sequences have been deleted so that the gene can no longer be transcribed, so-called gene knock-out studies (Capecchi, 1994). One of the first knock-out genes for behavior showed that knocking out a kinase gene interfered with learning a spatial task (Silva, Paylor, Wehner, & Tonegawa, 1992). There has been an explosion of research using targeted mutations in mice to study learning and memory, even though each experiment requires 2 or 3 years (Mayford & Kandel, 1999; Wahlsten, 1999). Newer techniques can produce more subtle changes that alter the genes regulation and lead to increases or decreases in the frequency with which the gene is transcribed. Techniques are even available to affect particular brain regions and to turn genes on and off at will. The approach is not without problems, however. Currently, there is no way to control the location of gene insertion in the mouse genome nor can the number of inserted copies of the gene be controlled, both of which can affect gene function. Another major approach, mutagenesis screening, is the opposite of targeting. Chemical mutagens are used to mutate genes at random, and thousands of mutated mice are screened for a wide variety of phenotypes (Nolan et al., 2000). The focus of this work is usually on single genes of major effect rather than QTLs with more subtle effects. Moreover, the necessity for screening thousands of mice makes it difficult to include intensive behavioral measures as needed, for example, to assess learning and memory. A different approach, called antisense DNA, circumvents some of these problems and does not require breeding. Antisense DNA is a DNA sequence that binds to a specific RNA sequence and thus prevents some of the RNA from being translated, which knocks down gene function. Injected in the brain, antisense DNA has the advantage of high temporal and spatial resolution (Ogawa & Pfaff, 1996). An early antisense knockdown study demonstrated the importance of the CRE-binding protein (CREB) gene in memory formation (Guzowski & McGaugh, 1997). Antisense DNA knockdowns have been shown to affect behavioral responses for dozens of drugs (Buck, Crabbe, & Belknap, 2000). The principal limitations of antisense technology currently are its unpredictable efficacy and a tendency to produce general toxicity.

Genomics and Postgenomics

As the advances from the Human Genome Project begin to be absorbed in DNA research on intelligence and other personality traits, optimism is warranted about finding more QTLs. The future of genetic research will involve a shift from finding genes to finding out how genes workthe pathways from genes to behavior. This research direction, generally referred to as functional genomics, is happening already in basic molecular biology as we approach the postgenomic era in which the complete human genome sequence and all functional variations in the genome sequence are identified (Plomin, DeFries, Craig, & McGuffin, 2003).

SPECIAL SECTION: 100 YEARS AFTER SPEARMAN (1904)

125

A more fundamental issue involves the interpretation of manipulating genes by knocking them down or out. Many genes, perhaps hundreds or thousands, provide the substrate for any behavior. Knocking out or knocking down the function of any one of these genes could affect a particular behavior, but this would not imply that the gene causes that behavior. An analogy is an automobile that requires hundreds of parts to work properly, but disabling any single part such as a spark plug or even a wheel will interfere with its functioning. Conversely, living systems have evolved redundancies that can compensate for malfunctionsit often happens that genes that are ostensibly crucial have little effect when knocked out. Moreover, the effect of a knockout is often very different in different inbred strains, suggesting the importance of other genes (Nadeau, 2001). Finally, showing that a gene affects behavior when the gene is knocked out or knocked down does not necessarily imply that naturally occurring variation in the gene is responsible for the heritability of that behavior.

Proteomics
Gene expression profiling assesses gene transcription as indexed by RNA. The next step toward functional genomics is to study the proteins that result from translation of RNA and their interactions. The phrase protein genomics led to the neologism proteomics. Just as there is no single human genome, there is no single human proteome. For understanding how individual differences in behavior are caused by DNA differences, the first step is to investigate the differences in protein function for which the key is their shape and complexes that they form with other proteins. Proteomics is much more difficult than genomics because, unlike the triplet code of DNA that governs the genome, there is no simple code for understanding the proteome. Moreover, there are many more proteins than genes for two reasons. First, it has been estimated that about half of all human genes are alternatively spliced into exons and introns and thus translated into different proteins (Venter et al., 2001). Second, after translation, proteins are modifiedit has been estimated that for each human gene three different modified proteins with different functions are produced (Banks et al., 2000).

Gene Expression Profiling

Genes are transcribed, or expressed, as their products are needed. As you read this sentence, you are creating new neurotransmitters by transcribing neurotransmitter genes. Gene expression can be indexed by the presence of RNA, which transcribes DNA, and then travels outside the nucleus to be translated into amino acid sequences, the building blocks of proteins. Microarrays are now available that can detect the expression of thousands of genes simultaneously. Such gene expression profiling was first used in cell lines to diagnose diseases on the basis of the profile of genes that are expressed in response to the disease (Golub et al., 1999) and to study the response to drugs (Iyer et al., 1999). Unlike DNA studies in which every cell in the body has the same DNA, gene expression studies depend on the tissue that is sampled. For personality traits, the brain is of course the critical tissue, which will make it difficult to apply this technology to the human species. However, gene expression profiling is being used widely in research on animal models, especially mice. Gene expression profiling comparing brain tissue before and after an eventlearning or stress, for example can point to genes whose expression is triggered by the event. Gene expression profiling can also be put to use to compare mice that differ genetically, such as mice bred for differential response to learning or stress, or inbred strains of mice that differ for many behaviors. It is also possible to combine these two approaches to study genotype-environment interaction. For example, a gene expression profiling study of more than 7,000 genes investigated gene expression in the hippocampus during ethanol withdrawal following chronic ethanol exposure for two inbred strains of mice (G. M. Daniels & Buck, 2002). Approximately 2% of genes in one strain and 1% of genes in the other strain were differentially expressed before and after withdrawal, and a few genes were identified that showed the largest differential response to withdrawal in the two strains. Similar results appear to be emerging for other manipulations, such as drug-induced seizures (Sandberg et al., 2000). Gene expression profiling may be useful in nominating QTLs and it will certainly be useful in understanding how QTLs associated with complex traits function. It is analogous to functional neuroimaging at the level of the gene.

Behavioral Genomics
Gene manipulation, gene expression profiling, and proteomics are examples of bottom-up molecular biological approaches to functional genomics. However, there are other levels of analysis at which we can understand how genes work. For example, one step up from proteomics is the analysis of molecular changes in the synapse, which is the focus of neurogenetics research on learning and memory (Grant, 2003). Pathways in the brain between genes and intelligence can be traced using neuroimaging techniques in the human species (Kosslyn & Plomin, 2001) and even more precise techniques can be used in the mouse (Crusio & Gerlai, 1999). As Spearman noted in 1927, ultimate understanding of g must needs come from the most profound and detailed direct study of the human brain in its purely physical and chemical aspects (p. 403). At the other end of the continuum from bottom-up approaches of molecular biology is a top-down level of analysis that considers the behavior of the whole organism. The phrase behavioral genomics has been suggested to emphasize the potential contribution of a top-down psychological level of analysis toward understanding how genes work (Plomin & Crabbe, 2000). For example, part of understanding how genes work is to understand how genetic effects interact and correlate with experience, how genetic effects on behavior contribute to change and continuity in development, and how genetic effects contribute to overlap between traits. As discussed earlier, these are issues central to quantitative genetic analyses. Behavioral genomic research using DNA will provide sharper scalpels to dissect these issues with greater precision. Table 1 describes examples of environmental, developmental, and multivariate findings mentioned earlier from quantitative genetic research on intelligence and poses questions they raise for behavioral genomic research on intelligence. Behavioral genomics will make important contributions toward understanding the functions of genes and DNA will open up new horizons for understanding behavior. Few personality researchers are likely to join the hunt for genes because it is difficult and expensive, but once genes are found, it is relatively easy and

126

PLOMIN AND SPINATH

Table 1 Examples of Quantitative Genetic Findings on Intelligence and the Questions They Raise for Behavioral Genomics
Issues Environmental GE correlation GE interaction Developmental Cross-sectional Longitudinal Multivariate Genetic g Links between normal and abnormal Quantitative genetic findings Associations between environmental measures and intelligence are often mediated genetically. Heritability can differ as a function of environment. Heritability increases with age. From age to age, genetic influence shows much continuity and some change. The same genes affect diverse cognitive abilities. MMR is the quantitative extreme of the same genes responsible for heritability throughout the distribution. Behavioral genomics questions To what extent do QTLs drive experience? Do QTL associations differ as a function of environment? Which QTL associations increase with age? To what extent do QTLs expressed early in life predict later behavior? Which QTLs are responsible for genetic overlap? Which QTLs associated with MMR are also associated with normal variation?

Note. GE genotype environment; QTLs quantitative trait loci; MMR mild mental retardation. From Behavioral Genetics in the Postgenomic Era (p. 533), by R. Plomin, J. C. Defries, I. W. Craig, and P. McGuffin, 2003, Washington, DC: American Psychological Association. Copyright 2003 by the American Psychological Association.

inexpensive to use them. DNA can be obtained painlessly and inexpensively from cheek swabs blood is not necessary. Cheek swabs yield enough DNA to genotype thousands of genes, and the cost of genotyping is surprisingly inexpensive. What has happened in the area of dementia in the elderly will be played out in many other areas of the behavioral sciences including personality. As noted earlier, the only known risk factor for late-onset Alzheimers dementia (LOAD) is APOE. Although the association between allele 4 of APOE and LOAD was reported a decade ago (Corder et al., 1993), it has already become routine in research on dementia to genotype subjects for APOE in order to ascertain whether the results differ for individuals with and without this genetic risk factor. For example, the association between APOE and dementia appears to interact with head injury, smoking, cholesterol level, and estrogen level (Williams, 2003). For these reasons, we predict that personality researchers will routinely collect DNA in their research and incorporate identified QTLs in their research, which will greatly enrich behavioral genomics.

hypertension, and obesity. Behavioral genetics will contribute a quantitative genetic and QTL perspective that shifts the focus of common disorders to dimensions of normal variation in which common disorders are viewed as the quantitative extreme of the same genetic and environmental factors that create variation throughout the distribution. This shift has begun in genetic research on psychopathology, the most active area of behavioral genetic research (McGuffin, Gottesman, & Owen, 2002). This shift is leading to renewed interest in personality as the source of normal variation. This integration is more than methodological and technological. Because DNA is the ultimate common denominator for research, postgenomic research on intelligence and personality will increasingly become integrated into the life sciences. The bottom-up approaches of molecular biology will eventually meet the top-down approaches of behavioral genomics in the brain.

References
Abecasis, G. R., Noguchi, E., Heinzmann, A., Traherne, J. A., Bhattacharyya, S., Leaves, N. I., et al. (2001). Extent and distribution of linkage disequilibrium in three genomic regions. American Journal of Human Genetics, 68, 191197. Baker, E., Hinton, L., Callen, D. F., Altree, M., Dobbie, A., Eyre, H. J., et al. (2002). Study of 250 children with idiopathic mental retardation reveals nine cryptic and diverse subtelomeric chromosome anomalies. American Journal of Medical Genetics, 107, 285293. Banks, R. E., Dunn, M. J., Hochstrasser, D. F., Sanchez, J. C., Blackstock, W., Pappin, D. J., & Selby, P. J. (2000). Proteomics: New perspectives, new biomedical opportunities. Lancet, 356, 1749 1756. Bayley, N. (1969). Manual for the Bayley Scales of Infant Development. New York: Psychological Corporation. Benjamin, J., Ebstein, R., & Belmaker, R. H. (2002). Molecular genetics of human personality. Washington, DC: American Psychiatric Press.

Conclusions
The future for genetic research in intelligence and other areas of personality looks brighter than ever in the dawn of the postgenomic era. The genetics of such complex traits will be swept along in the wake of the Human Genome Project as it increasingly provides the tools needed for the genetic analysis of complex traits. The most exciting prospect is the integration of quantitative genetics, molecular genetics, and functional genomics in a new focus on behavioral genomics (Plomin et al., 2003). Behavioral genetics will profit from as well as contribute to this integration. It will profit from the advances coming from intense molecular genetic research on common complex medical disorders such as diabetes,

SPECIAL SECTION: 100 YEARS AFTER SPEARMAN (1904) Benjamin, J., Li, L., Patterson, C., Greenburg, B. D., Murphy, D. L., & Hamer, D. H. (1996). Population and familial association between the D4 dopamine receptor gene and measures of novelty seeking. Nature Genetics, 12, 81 84. Bock, G., Goode, J. A., & Webb, K. (Eds.). (2000). The nature of intelligence. London: Novartis Foundation. Bouchard, T. J., Jr., Lykken, D. T., McGue, M., Segal, N. L., & Tellegen, A. (1990, October 12). Sources of human psychological differences: The Minnesota study of twins reared apart. Science, 250, 223228. Bouchard, T. J., Jr., & McGue, M. (1981, May 29). Familial studies of intelligence: A review. Science, 212, 10551059. Bradley, R. H., Convyn, R. F., Burchinal, M., McAdoo, H. P., & Coll, C. G. (2001). The home environments of children in the United States part II: Relations with behavioral development through age thirteen. Child Development, 72, 1868 1886. Braungart, J. M., Fulker, D. W., & Plomin, R. (1992). Genetic mediation of the home environment during infancy: A sibling adoption study of the HOME. Developmental Psychology, 28, 1048 1055. Brett, D., Pospisil, H., Valcarcel, J., Reich, J., & Bork, P. (2002). Alternative splicing and genome complexity. Nature Genetics, 30, 29 30. Brody, N. (1992). Intelligence (2nd ed.). New York: Academic Press. Buck, K. J., Crabbe, J. C., & Belknap, J. K. (2000). Alcohol and other abused drugs. In D. W. Pfaff, W. H. Berrettini, T. H. Joh, & S. C. Maxson (Eds.), Genetic influences on neural and behavioral functions (pp. 159 183). Boca Raton, FL: CRC Press. Burks, B. (1928). The relative influence of nature and nurture upon mental development: A comparative study on foster parent-foster child resemblance. Yearbook of the National Society for the Study of Education, Part 1, 27, 219 316. Burt, C. (1966). The genetic determination of differences in intelligence. British Journal of Psychology, 57, 137153. Capecchi, M. R. (1994). Targeted gene replacement. Scientific American, 270, 5259. Cardon, L. R., & Bell, J. (2001). Association study designs for complex diseases. Nature Genetics, 2, 9199. Cardon, L. R., Smith, S. D., Fulker, D. W., Kimberling, W. J., Pennington, B. F., & DeFries, J. C. (1994, October 14). Quantitative trait locus for reading disability on chromosome 6. Science, 266, 276 279. Carroll, J. B. (1993). Human cognitive abilities. New York: Cambridge University Press. Chipuer, H. M., & Plomin, R. (1992). Using siblings to identify shared and non-shared HOME items. British Journal of Developmental Psychology, 10, 165178. Chipuer, H. M., Rovine, M. J., & Plomin, R. (1990). LISREL modeling: Genetic and environmental influences on IQ revisited. Intelligence, 14, 1129. Chorney, M. J., Chorney, K., Seese, N., Owen, M. J., Daniels, J., McGuffin, P., et al. (1998). A quantitative trait locus (QTL) associated with cognitive ability in children. Psychological Science, 9, 1 8. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Corder, E. H., Saunders, A. M., Strittmatter, W. J., Schmechel, D. E., Gaskell, P. C., Small, G. W., et al. (1993, August 13). Gene dose of apolipoprotein E type 4 allele and the risk of Alzheimers disease in late onset families. Science, 261, 921923. Craig, I., & McClay, J. (2003). The role of molecular genetics in the postgenomics era. In R. Plomin, J. C. DeFries, I. W. Craig, & P. McGuffin (Eds.), Behavioral genetics in the postgenomic era (pp. 19 40). Washington, DC: American Psychological Association. Crusio, W. E., & Gerlai, R. T. (Eds.). (1999). Handbook of moleculargenetic techniques for brain and behavior research. Amsterdam: Elsevier. Daniels, G. M., & Buck, K. J. (2002). Expression profiling identifies

127

strain-specific changes associated with ethanol withdrawal in mice. Genes, Brain and Behavior, 1, 35 45. Daniels, J., Holmans, P., Plomin, R., McGuffin, P., & Owen, M. J. (1998). A simple method for analyzing microsatellite allele image patterns generated from DNA pools and its application to allelic association studies. American Journal of Human Genetics, 62, 1189 1197. Deary, I. J. (2000). Looking down on human intelligence: From psychometrics to the brain. Oxford, England: Oxford University Press. Deary, I. J. (2001). Human intelligence differences: Towards a combined experimental-differential approach. Trends in Cognitive Science, 5, 164 170. DeFries, J. C., & Fulker, D. W. (1985). Multiple regression analysis of twin data. Behavior Genetics, 15, 467 473. DeFries, J. C., & Fulker, D. W. (1988). Multiple regression analysis of twin data: Etiology of deviant scores versus individual differences. Acta Geneticae Medicae et Gemellologicae, 37, 205216. Devlin, B., Daniels, M., & Roeder, K. (1997). The heritability of IQ. Nature, 388, 468 471. Ebstein, R. P., Novick, O., Umansky, R., Priel, B., Osher, Y., Blaine, D., et al. (1996). Dopamine D4 receptor (D4DR) exon III polymorphism associated with the human personality trait novelty-seeking. Nature Genetics, 12, 78 80. Eddy, S. R. (2001). Non-coding RNA genes and the modern RNA world. Nature Reviews Genetics, 2, 919 929. Egan, M. F., Goldberg, T. E., Kolachana, B. S., Callicott, J. H., Mazzanti, C. M., Straub, R. E., et al. (2001). Effect of COMT Val108/158 Met genotype on frontal lobe function and risk for schizophrenia. Proceedings of the National Academy of Sciences of the United States of America, 98, 6917 6922. Egeland, J. A., Gerhard, D. S., Pauls, D. L., Sussex, J. N., Kidd, K. K., Allen, C. R., et al. (1987). Bipolar affective disorders linked to DNA markers on chromosome 11. Nature, 325, 783787. Erlenmeyer-Kimling, L., & Jarvik, L. F. (1963, December 13). Genetics and intelligence: A review. Science, 142, 14771479. Falconer, D. S. (1965). The inheritance of liability to certain diseases estimated from the incidence among relatives. Annals of Human Genetics, 29, 5176. Faraone, S. V., Doyle, A. E., Mick, E., & Biederman, J. (2001). Metaanalysis of the association between the 7-repeat allele of the dopamine D(4) receptor gene and attention deficit hyperactivity disorder. American Journal of Psychiatry, 158, 10521057. Farrer, L. A., Cupples, L. A., Haines, J. L., Hyman, B., Kukull, W. A., Mayeux, R., et al. (1997). Effects of age, sex, and ethnicity on the association between apolipoprotein E genotype and Alzheimer disease. A meta-analysis. APOE and Alzheimer Disease Meta Analysis Consortium. Journal of the American Medical Association, 278, 1349 1356. Finkel, D., Pedersen, N. L., McGue, M., & McClearn, G. E. (1995). Heritability of cognitive abilities in adult twins: Comparison of Minnesota and Swedish data. Behavior Genetics, 25, 321 431. Fisher, P. J., Turic, D., McGuffin, P., Asherson, P. J., Ball, D. M., Craig, I. W., et al. (1999). DNA pooling identifies QTLs for general cognitive ability in children on chromosome 4. Human Molecular Genetics, 8, 915922. Fisher, S. (2003). Isolation of the genetic factors underlying speech and language disorders. In R. Plomin, J. C. DeFries, I. W. Craig, & P. McGuffin (Eds.), Behavioral genetics in the postgenomic era (pp. 205 226). Washington, DC: American Psychological Association. Freeman, F. N., Holzinger, K. J., & Mitchell, B. (1928). The influence of environment on the intelligence, school achievement, and conduct of foster children. Yearbook of the National Society for the Study of Education, 27, 103217. Fulker, D. W., Cherny, S. S., & Cardon, L. R. (1993). Continuity and change in cognitive development. In R. Plomin & G. E. McClearn

128

PLOMIN AND SPINATH (Eds.), Psychiatric neuroimaging research: Contemporary strategies (pp. 491515). Washington DC: American Psychiatric Press. Kruglyak, L. (1999). Prospects for whole-genome linkage disequilibrium mapping of common disease genes. Nature Genetics, 22, 139 144. Loehlin, J. C. (1989). Partitioning environmental and genetic contributions to behavioral development. American Psychologist, 44, 12851292. Loehlin, J. C. (1992). Genes and environment in personality development. Newbury Park, CA: Sage. Loehlin, J. C., Horn, J. M., & Willerman, L. (1989). Modeling IQ change: Evidence from the Texas Adoption Project. Child Development, 60, 9931004. Mattick, J. S. (2001). Non-coding RNAs: The architects of eukaryotic complexity. EMBO Reports, 2, 986 991. Mayford, M., & Kandel, E. R. (1999). Genetic approaches to memory storage. Trends in Genetics, 15, 463 470. McClearn, G. E. (1963). The inheritance of behavior. In L. J. Postman (Ed.), Psychology in the making (pp. 144 252). New York: Knopf. McCrae, R. R. & Costa, P. T., Jr. (1997). Conceptions and correlates of Openness to Experience. In R. Hogan, J. Johnson, & S. Briggs (Eds.), Handbook of personality psychology (pp. 825 847). San Diego, CA: Academic Press. McGue, M. (1997). The democracy of the genes. Nature, 388, 417 418. McGue, M., Bouchard, T. J., Jr., Iacono, W. G., & Lykken, D. T. (1993). Behavioral genetics of cognitive ability: A life-span perspective. In R. Plomin & G. E. McClearn (Eds.), Nature, nurture, and psychology (pp. 59 76). Washington, DC: American Psychological Association. McGuffin, P., Gottesman, I. I., & Owen, M. J. (2002). Psychiatric genetics and genomics. Oxford, England: Oxford University Press. Merriman, C. (1924). The intellectual resemblance of twins. Psychological Monographs, 33, 158. Nadeau, J. H. (2001). Modifier genes in mice and humans. Nature Reviews Genetics, 2, 165174. Neale, M. C., Boker, S. M., Xie, G., & Maes, H. (1999). Mx: Statistical modeling (5th ed.). Richmond: Virginia Commonwealth University, Department of Psychiatry. Neisser, U., Boodoo, G., Bouchard, T. J., Jr., Boykin, A. W., Brody, N., Ceci, S. J., et al. (1996). Intelligence: Knowns and unknowns. American Psychologist, 51, 77101. Nichols, P. L. (1984). Familial mental retardation. Behavior Genetics, 14, 161170. Nolan, P. M., Peters, J., Strivens, M., Rogers, D., Hagan, J., Spurr, N., et al. (2000). A systematic, genome-wide, phenotype-driven mutagenesis programme for gene function studies in the mouse. Nature Genetics, 25, 440 443. Norton, N., Williams, N. M., Williams, H. J., Spurlock, G., Kirov, G., Morris, D. W., et al. (2002). Universal, robust, highly quantitative SNP allele frequency measurement in DNA pools. Human Genetics, 110, 471 478. Ogawa, S., & Pfaff, D. W. (1996). Application of antisense DNA method for the study of molecular bases of brain function and behavior. Behavior Genetics, 26, 279 292. Pedersen, N. L., McClearn, G. E., Plomin, R., & Nesselroade, J. R. (1992). Effects of early rearing environment on twin similarity in the last half of the life span. British Journal of Developmental Psychology, 10, 255 267. Pennington, B. C., Filipek, P. A., Lefly, D., Chhabildas, N., Kennedy, D. N., Simon, J. H., et al. (2000). A twin MRI study of size variations in the human brain. Journal of Cognitive Neuroscience, 12, 223232. Petrill, S. A. (1997). Molarity versus modularity of cognitive functioning? A behavioral genetic perspective. Current Directions in Psychological Science, 6, 96 99. Plomin, R. (1986). Development, genetics, and psychology. Hillsdale, NJ: Erlbaum. Plomin, R. (1988). The nature and nurture of cognitive abilities. In R. J.

(Eds.), Nature, nurture, and psychology (pp. 7797). Washington, DC: American Psychological Association. Galsworthy, M., Paya-Cano, J. L., Monleon, S., & Plomin, R. (2002). General cognitive ability (g) and potential confounds in heterogeneous stock (HS) mice. Genes, Brain and Behavior, 1, 88 95. Galton, F. (1865). Heredity, talent and character. Macmillans Magazine, 12, 157166; 318 327. Galton, F. (1869). Heredity, genius: An enquiry into its laws and consequences. London: Macmillan. Galton, F. (1876). The history of twins as a criterion of the relative powers of nature and nurture. Royal Anthropological Institute of Great Britain and Ireland Journal, 6, 391 406. Galton, F. (1883). Inquiries into human faculty and its development. London: Macmillan. Goldstein, D. B. (2001). Islands of linkage disequilibrium. Nature Genetics, 29, 109 111. Golub, T. R., Slonim, D. K., Tamayo, P., Huard, C., Gaasenbeek, M., Mesirov, J. P., et al. (1999, October 15). Molecular classification of cancer: Class discovery and class prediction by gene expression monitoring. Science, 286, 531537. Grant, S. G. N. (2003). An integrative neuroscience program linking genes to cognition and disease. In R. Plomin, J. C. DeFries, I. W. Craig, & P. McGuffin (Eds.), Behavioral genetics in the postgenomic era (pp. 123 138). Washington, DC: American Psychological Association. Gusella, J. F., Wexler, N. S., Conneally, P. M., Naylor, S. L., Anderson, M. A., & Tanzi, R. E. (1983). A polymorphic DNA marker genetically linked to Huntingtons disease. Nature, 306, 234 238. Guzowski, J. F., & McGaugh, J. L. (1997). Antisense oligodeoxynucleotide-mediated disruption of hippocampal cAMP response element binding protein levels impairs consolidation of memory for water maze training. Proceedings of the National Academy of Sciences of the United States of America, 94, 26932698. Harris, J. R. (1998). The nurture assumption: Why children turn out the way they do. New York: Free Press. Herrnstein, R. J., & Murray, C. (1994). The bell curve: Intelligence and class structure in American life. New York: Free Press. Hill, L., Chorney, M. J., Chorney, K., Craig, I. W., Fisher, P., Owen, M. J., et al. (1999). IGF2R and cognitive ability. Molecular Psychiatry, 4, S108. Hill, L., Chorney, M. C., & Plomin, R. (2002). A quantitative trait locus (not) associated with cognitive ability? Psychological Science, 13, 561 562. Hill, L., Craig, I. W., Ball, D. M., Eley, T. C., Ninomiya, T., Fisher, P. J., et al. (1999). DNA pooling and dense marker maps: A systematic search for genes for cognitive ability. NeuroReport, 10, 843 848. International Human Genome Sequencing Consortium. (2001). Initial sequencing and analysis of the human genome. Nature, 409, 860 921. Iyer, V. R., Eisen, M. B., Ross, D. T., Schuler, G., Moore, T., Lee, J. C. F., et al. (1999, January). The transcriptional program in the response of human fibroblasts to serum. Science, 283, 83 87. Jensen, A. R. (1969). How much can we boost IQ and scholastic achievement? Harvard Educational Review, 39, 1123. Jensen, A. R. (1998). The g factor: The science of mental ability. Westport, CT: Praeger Publishers. Kamin, L. J. (1974). The science and politics of IQ. Potomac, MD: Erlbaum. Kendler, K. S., & Eaves, L. J. (1986). Models for the joint effects of genotype and environment on liability to psychiatric illness. American Journal of Psychiatry, 143, 279 289. Knight, S. J. L., Regan, R., Nicod, A., Horsley, S. W., Kearney, L., Homfray, et al. (1999). Subtle chromosomal rearrangements in children with unexplained mental retardation. Lancet, 354, 1676 1681. Kosslyn, S., & Plomin, R. (2001). Towards a neuro-cognitive genetics: Goals and issues. In D. Dougherty, S. L. Rauch, & J. F. Rosenbaum

SPECIAL SECTION: 100 YEARS AFTER SPEARMAN (1904) Sternberg (Ed.), Advances in the psychology of human intelligence, Vol. 4 (pp. 133). Hillsdale, NJ: Erlbaum. Plomin, R. (1994). Genetics and experience: The interplay between nature and nurture. Newbury Park, CA: Sage. Plomin, R. (1999). Genetic research on general cognitive ability as a model for mild mental retardation. International Review of Psychiatry, 11, 34 36. Plomin, R. (2001). The genetics of g in human and mouse. Nature Reviews Neuroscience, 2, 136 141. Plomin, R., & Bergeman, C. S. (1991). The nature of nurture: Genetic influences on environmental measures. Behavioral and Brain Sciences, 14, 373 427. Plomin, R., & Caspi, A. (1998). DNA and personality. European Journal of Personality, 12, 387 407. Plomin, R., & Crabbe, J. C. (2000). DNA. Psychological Bulletin, 126, 806 828. Plomin, R., & Daniels, D. (1987). Why are children in the same family so different from each other? Behavioral and Brain Sciences, 10, 116. Plomin, R. & DeFries, J. C. (1985). Origins of individual differences in infancy. Orlando, FL: Academic Press. Plomin, R., & DeFries, J. C. (1998, May). Genetics of cognitive abilities and disabilities. Scientific American, 62 69. Plomin, R., DeFries, J. C., Craig, I. W., & McGuffin, P. (2003). Behavioral genetics in the postgenomic era. Washington, DC: American Psychological Association. Plomin, R., DeFries, J. C., McClearn, G. E., & McGuffin, P. (2001). Behavioral genetics (4th ed.). New York: Worth Publishers. Plomin, R., Fulker, D. W., Corley, R., & DeFries, J. C. (1997). Nature, nurture and cognitive development from 1 to 16 years: A parent offspring adoption study. Psychological Science, 8, 442 447. Plomin, R., Hill, L., Craig, I., McGuffin, P., Purcell, S., Sham, P., et al. (2001). A genome-wide scan of 1842 DNA markers for allelic associations with general cognitive ability: A five-stage design using DNA pooling. Behavior Genetics, 31, 497509. Plomin, R., McClearn, G. E., Smith, D. L., Skuder, P., Vignetti, S., Chorney, M. J., et al. (1995). Allelic associations between 100 DNA markers and high versus low IQ. Intelligence, 21, 31 48. Plomin, R., Owen, M. J., & McGuffin, P. (1994, June 17). The genetic basis of complex human behaviors. Science, 264, 17331739. Plomin, R., Pedersen, N. L., Lichtenstein, P., & McClearn, G. E. (1994). Variability and stability in cognitive abilities are largely genetic later in life. Behavior Genetics, 24, 207215. Plomin, R., & Price, T. (2003). Genetics and intelligence. In N. Colangelo & G. A. Davis (Eds.), Handbook of gifted education (3rd ed., pp. 113123). Boston: Allyn & Bacon. Plomin, R., & Spinath, F. M. (2002). Genetics and general cognitive ability (g). Trends in Cognitive Science, 6, 169 176. Prolo, P. & Licinio, J. (2002). D4DR and novelty seeking. In J. Benjamin, R. Ebstein, & R. H. Belmaker (Eds.), Molecular genetics and human personality (pp. 91107). New York: American Psychiatric Press. Purcell, S., & Sham, P. C. (2003). A model-fitting implementation of the DeFries-Fulker model for selected twin data. Behavior Genetics, 33, 271278. Reed, E. W., & Reed, S. C. (1965). Mental retardation: A family study. Philadelphia: Saunders. Reich, D. E., Cargill, M., Bolk, S., Ireland, J., Sabeti, P. C., Richter, D. J., et al. (2001). Linkage disequilibrium in the human genome. Nature, 411, 199 204. Risch, N. J. (2000). Searching for genetic determinants in the new millennium. Nature, 405, 847 856. Risch, N., & Merikangas, K. R. (1996, September 13). The future of genetic studies of complex human diseases. Science, 273, 1516 1517. Ronald, A., Spinath, F. M., & Plomin, R. (2002). High cognitive ability in

129

early childhood: An etiological study using twins. High Ability Studies, 13, 103114. Sandberg, R., Yasuda, R., Pankratz, D. G., Carter, T. A., Del Rio, J. A., Wodicka, L., et al. (2000). Regional and strain-specific gene expression mapping in the adult mouse brain. Proceedings of the National Academy of Sciences of the United States of America, 97, 11038 11043. Sherrington, R., Brynjolfsson, J., Petursson, H., Potter, M., Dudleston, K., Barraclough, B., et al. (1988). Localisation of susceptibility locus for schizophrenia on chromosome 5. Nature, 336, 164 167. Silva, A. J., Paylor, R., Wehner, J. M., & Tonegawa, S. (1992, July 10). Impaired spatial learning in -calcium-calmodulin kinase mutant mice. Science, 257, 206 211. Silver, L. M. (1995). Mouse genetics: Concepts and applications. Oxford, England: Oxford University Press. Snyderman, M. & Rothman, S. (1987). Survey of expert opinion on intelligence and aptitude testing. American Psychologist, 42, 137144. Snyderman, M., & Rothman, S. (1988). The IQ controversy, the media and publication. New Brunswick, NJ: Transaction Publishers. Spearman, C. (1904). General intelligence, objectively determined and measured. American Journal of Psychology, 15, 201292. Spearman, C. (1927). The abilities of man: Their nature and measurement. New York: Macmillan. Spinath, F. M., Harlaar, N., Ronald, A., & Plomin, R. (in press). Substantial genetic influence on mild mental impairment in early childhood. American Journal of Mental Retardation. Spinath, F. M., Ronald, A., Harlaar, N., Price, T. S., & Plomin, R. (2003). Phenotypic g early in life: On the etiology of general cognitive ability in a large population sample of twin children aged 2 to 4 years. Intelligence, 31, 195210. Thapar, A. (2003). Attention deficit hyperactivity disorder: New genetic findings, new directions. In R. Plomin, J. C. DeFries, I. W. Craig, & P. McGuffin (Eds.), Behavioral genetics in the postgenomic era (pp. 445 462). Washington, DC: American Psychological Association. Theis, S. V. S. (1924). How foster children turn out (Publication No. 165). New York: State Charities Aid Association. Thompson, P. M., Cannon, T. D., Narr, K. L., van Erp, T., Poutanen, V. P., Huttunen, M., et al. (2001). Genetic influences on brain structure. Nature Neuroscience, 4, 12531258. Turic, D., Fisher, P. J., Plomin, R., & Owen, M. J. (2001). No association between apolipoprotein E polymorphisms and general cognitive ability in children. Neuroscience Letters, 299, 97100. Venter, J. C., Adams, M. D., Myers, E. W., Li, P. W., Mural, R. J., Sutton, G. G., et al. (2001, February). The sequence of the human genome. Science, 291, 1304 1351. Wahlsten, D. (1999). Single-gene influences on brain and behavior. Annual Review of Psychology, 50, 599 624. Wickelgren, I. (1998, April 24). Tracking insulin to the mind. Science, 280, 517519. Willcutt, E. G., DeFries, J. C., Pennington, B. F., Smith, S. D., Cardon, L. R., & Olson, R. K. (2003). Comorbid reading difficulties and ADHD. In R. Plomin, J. C. DeFries, I. C. Craig, & P. McGuffin (Eds.), Behavioral genetics in the postgenomic era (pp. 227246). Washington, DC: American Psychological Association. Williams, J. (2003). Dementia. In R. Plomin, J. C. DeFries, I. W. Craig, & P. McGuffin (Eds.), Behavioral genetics in the postgenomic era (pp. 503527). Washington, DC: American Psychological Association. Zechner, U., Wilda, M., Kehrer-Sawatzki, H., Vogel, W., Fundele, R., & Hameister, H. (2001). A high density of X-linked genes for general cognitive ability: A runaway process shaping human evolution? Trends in Genetics, 17, 697701.

Received February 18, 2002 Revision received August 26, 2002 Accepted October 3, 2002