ps418

In an article appearing in the february issue of the journal Current Anthropology, Paul Ehrlich and Marcus Feldman advance a number of criticisms of evolutionary psychology and behavioral genetics. Ehrlich and Feldman’s article makes some valid and uncontroversial points; however, as critical commentary on their article from a number of researchers makes clear, they largely do battle with an army of strawmen (e.g. Hauser and Wrangham, p. 97). Here I comment on a few of Ehrlich and Feldman's assertions, particularly those relating to behavior genetics and the heritability of brain structure in humans.


Heritability of Human Traits

Ehrlich and Feldman assert (p. 90) that heritability estimates can not be assigned to human behavioral traits because "[a]n accurate measure of heritability requires that parents and offspring be raised in identical environments." First, it is not the case that measurement of heritability requires parent-offspring comparisons. Heritability is the proportion of phenotypic variance that is attributable to genotypic variance. Parent-offspring is one level of genetic relatedness that can be used to calculate heritability, but not the only one. In fact, several recent behavior genetic analyses incorporate several levels of genetic relatedness (e.g. half-sibs, parent-offspring, full siblings, monozygotic twins). Second, according to Ehrlich and Feldman's requirement, it would be illicit to assign any measure of heritability to any human trait, not just psychological, but physiological and anthropometric phenotypes as well, such as Alzheimer's disease or height or fingerprint ridge count or pitch recognition (see Drayna et al., 2001). Third, though it always remains a possibility that the results of this or that study has been confounded by an uncontrolled environmental variable, it is possible to test for the effects of individual confounding variables (e.g. the chorion in twin studies, or parental knowledge of zygosity). Fourth, Ehrlich and Feldman miss the larger point: whether or not we can determine a precise heritability estimate that applies at all times in all populations, the evidence from twin and adoption studies make it abundantly clear that heritability for many behavioral phenotypes is significant. And though it is not possible to study humans raised in identical environments (nor is it clear why this would be more informative!), it is nonetheless significant that heritability estimates for broad personality traits and cognitive ability have proven to replicate very well across time, populations, assessment methods and research designs.

Ehrlich and Feldman (p. 91) quote the rhetorical question of Goldberger and Kamin: “What conceivable purpose is served by the flood of heritability estimates generated by these studies?” The answer is both simple and obvious: to inform discussion about the sources of individual differences with respect to certain phenotypes in humans. For instance, it would foolish to suggest that homonid brain expansion occurred via sexual selection for brain size if brain size did not possess significant heritability, because only traits with signficant heritability will respond to selection. Likewise it would be a waste of time and resources to search for genes associated with a trait that shows insignficant heritability. On the other hand, Ehrlich and Feldman are correct to point out that a high heritability does not necessarily mean unmalleable, as is sometimes assumed by both proponents and critics of behavior genetics. Individual differences in height are almost completely the due to genetic differences, yet mean population height has increased significantly over the past century. Malleability of a behavior is a seperate, empirical question. Though a high heritability does imply that genetic differences rather than environmental differences are the major source of inividual differences with respect to that phenotype, this says something about what is, not about what can be.

Ehrlich and Feldman write (p. 102) that "[n]either the length nor the width of a rectangle controls the area but rather an interaction between the two, just as we say that 'every aspect of a person’s phenome is a product of interaction between genome and environment.'” This analogy is based on a common but fundamental misunderstanding of behavior genetics. Behavior genetics does not require the incoherent assumption that a behavior can be seperated into environmental and genetic components. Behavior geneticists know as well as anyone that “every aspect of a person’s phenome is a product of interaction between genome and environment." What behavior genetics does attempt to untangle are sources of variance in behavioral traits in a population. These are entirely different concepts. Though it may indeed be meaningless to ask "what is the genetic component of that behavior?," it is, in the words of Dobzhansky (1964, p. 55), "far from meaningless" to ask "[t]o what extent are the differences observed among people conditioned by the differences of their genotypes and by differences between the environments in which people were born . . . and brought up?"

Returning to the rectangle analogy, suppose we had a “population” of 4 rectangles. Their widths are all precisely 1 meter, while their lengths are 2, 3, 4, and 5 meters, respectively. The phenotype "area" for every rectangle is a function of both length and width. You cannot say x proportion of area of any rectangle is "caused" by length or width. Yet, you can say that, in this population of rectangles, variance in area is completely accounted for by variance in length, while none of the variance in area is accounted for variation in width. Likewise, to use a real-world example, a person's height is the product of developmental interactions between genes and environments. It would be meaningless to say that your lower 5 feet of height is genetic while the upper 10 inches is environmental. But it would be "far from meaningless" to say that across a wide range of (western) environments, individual differences in height are due primarily to genetic differences.

Ehrlich and Feldman note (p. 90) that "broad-sense heritability has no predictive value and indeed cannot be legitimately used in the human behavioral context to predict anything. It has, however, been widely misinterpreted as diagnostic of the underlying causes of variation." First, contrary to Ehrlich and Feldman, BSH is predictive, but only for the small group of genetically identical individuals, who share both additive and nonadditive (dominance and epistatic) genetic components. For instance, BSH of a trait predicts very strongly the mean intraclass correlations for measures of that trait in reared-apart MZ twins. No, it will not by itself allow you to predict population response to selection, because only the additive component (narrow heritability) will respond predictably to selection. Second, contrary to Ehrlich and Feldman, BSH is in fact a measure of the degree to which individual differences are the result of genetic differences. For instance, Otto (2001, p. 7653), who Ehrlich cites elsewhere in his article, states that "broad-sense heritability estimates the degree to which differences among individuals are genetically based," even though "not all genetically based differences can be passed from parents to offspring." It is unclear what Ehrlich and Feldman mean when they say that BSH has been "misinterpreted."


Gene x Environment Interactions

Ehrlich and Feldman (p. 91) assert that the "even if a behavior had a high degree of heritability in one environment, a small environmental alteration could totally change that behavior. The literature on quantitative traits in plants, insects, and animals is replete with experiments that show the sensitivity of measured heritability to changes in the environment." Absolutely! The prospect of identifying such strong gene x environment interactions is one of the primary reasons one would want to conduct behavior genetic analyses of human behavior in the first place. Psychologists, sociologists, politicians, and everyone else would like to know what are those small environmental alterations that "totally change that behavior"?

Gene x environment interactions can be thought of as differential individual response to environmental influences due to genotypic differences. Examples of such interactions are common in human disease. A widely known example is the case of phenylketonuria and the alleles associated with that disease. For most human genotypes - those with normal phenylalinine hydroxylase genes, the development of cognitive ability is relatively insensitive to dietary variations in phenylalinine, and will not develop mental retardation whether they are on a high or low phenylalynine diet. But there is a strong environment x gene interaction in the development of cognitive abilities in those with a mutated PAH gene, such that those who possess a mutated PAH gene will strongly adversely affected by high phenylalinine "environment," but will develop normal cognitive ability in a low phenylaline "environment." As another example from medical genetics, Mayeax et al (1995) found that there is no increase in risk for Alzheimer's associated with head injury in the absence of the gene Apo-E-4, that there was a 2 fold increase in risk associated with possession of the Apo-E-4 allele but not head injury, and a 10 fold increase in risk for Alzheimer's associated with head injuires in thosepossessing the Apo-E-4 allele. It would be a signficant acheivement to find the behavioral equivalents of the low-phenylalinine diet that would allow us to prevent the emergence of pathological behavioral or psychological phenotypes (e.g. autism, schizophrenia, mental retardation).
However, though strong gene x environment interactions are common for quantitative traits in plants and for disease in humans, they are fairly rare for quantitative behavioral traits in mammals. Rowe and Jacobson (2001, p. 28) note that "surprisingly few" signficant interactions have been found for quantitative behavioral traits in humans, and that "in most cases, additive models, which do not allow for G x E interactions, fit data extremely well" (see also McCall, 1991).

However, some examples of signficant gene x environment interactions have been found, particularly for the development of psychopathology. For instance, Cadoret et al (1985) found evidence for gene x environment interaction in the development of antisocial behavior in a relatively large sample (n=367) of adopted adolescents. One group of adoptees was considerd to be at "genetic risk" for developing such problems, as crudely indexed by the precense of similar problems in their biological parents, while the other group was not considered at genetic risk, as crudely indexed by the absence of antisocial behavior in their parents. Adoptive environments were considered favorable or advere based on the precense or absence of antisocial behavior, anxiety/depression, drug abuse, and marital problems among adoptive parents. Cadoret et al found that the group that was not at genetic risk developed did not develop antisocial behaviors, whether the adoptive family conditions were adverse or favorable, and that the group at genetic risk had only a slightly elevated amount of antisocial behavior problems when the adoptive home environments were favorable. However, of the genetically at risk group reared in adverse adoptive homes, there was a substantial effect.
Caspi et al (2002) provide evidence for gene x environment interaction in the development of antisocial behavior in a large sample (n=1037) of individuals that were assessed longitudinally at 3,5,7,9,11,13,15,18, 21, and 26 years. Approximately 8% of the sample experienced severe maltreatment, and another 28% probable maltreatment. The participants were assessed for antisocial behavior at various stages, using 4 different measures [DSM III adolescent conduct disorder, criminal conviction records, personality assessment, peer rating] and a composite index, and were genotyped with respect to a particular MAOA [Monoamine oxidase A] promoter polymorphism known to reduce MAOA expression. The results suggest that, of those who suffered childhood maltreatment, those with the low-MAOA promoter polymorphism were much more likely to develop antisocial behavior problem than those with high-MAOA genotype. The low-MAOA /maltreated group accounted for only 12% of the sample, but accounted for 44% of the violent crime convictions. 85% of the low-MAOA/severely maltreated group developed some form of antisocial behaviors. Among those with normal MAOA expression, childhood maltreatment did not confer significant risk for later conduct disorder. Basically, MAOA genotype appears to contributes very significant risk for development of antisocial behavior among those who are maltreated, but not among those who are not maltreated. For a review of research on gene x environment interactions in relation to psychopathology, see Rutter and Silberg (2002). Again, the point being made is that although every aspect of human behavior and psychology is the product of developmental interactions between genes and environments, it is both meaningful and possible to examine the extent which individual differences are "conditioned by the differences of their genotypes and by differences between the environments in which people were born . . . and brought up".

Twin Study Methodology
Ehrlich and Feldman assert (p. 90) that the statistical assumptions underlying twin studies have only recently been examined, and cite the meta-analysis of Devlin et al (1997) as having demonstrated that "omission of a contribution from the shared prenatal environment of twins also leads to elevated estimates of heritability." Devlin et al's (1997) meta-analysis incorporated 212 previous studies, and deduced broad and narrow heritabilities of IQ of approximately 0.5 and 0.34, respectively. Even taking Devlin et al's analysis at face value, the evidence for genetic influences on IQ variability remains highly significant. Nevertheless, there are several problems with that analysis which are not mentioned by Ehrlich and Feldman. First, even though they acknowledge that IQ heritability may increase with age, they attempt to minimize this evidence. If they had only included adults in their analysis, heritability would have been even higher. For instance, a panel of experts convened by the American Psychological Association concluded that the best estimate for heritability of IQ in adults is approximately 0.75 (Neisser et al., 1996, p. 85). This increase in heritability with age was well-established prior to 1997, and has since been repeatedly confirmed in longitudinal studies (e.g. Bartels et al., 2002). Secondly, the existing empirical research on the contribution of maternal environment to IQ similarity in twins does not support Devlin et al's (1997) thesis that such contributions are significant or have lead to inflated heritability estimates. McGue (1997), in his commentary on Devlin et al (1997) which appeared in the same issue of Nature, pointed out that "[c]aution is certainly warranted" in accepting these results too readily, because in "large scale studies where pre- or perinatal influences on IQ have been assessed directly, little evidence for any strong effect has been found." This is fully supported by more recent research comparing monochorionic MZ twins to dichorionic MZ twins. For instance, Jacobs et al. (2001) reported small but significant differences on 2 out of 15 WISC-R subscales (arithmetic and vocabulary) between monochorionic and dichorionic twins, accounting for about 10 and14% of the score variance on those two scales. There was no significant difference between the groups for 'total IQ' or the other 13 scales, which of course is hard to reconcile with Devlin et al's analysis that shared prenatal environment contributes significantly to IQ covariance or results in heritability overestimates.


Ehrlich and Feldman point out that "[w]e might act like our parents because they gave us our genes; however, as Richard Lewontin pointed out, “in the United States, the highest correlations between parent and offspring for any social traits are for religious sect and political party. Only the most vulgar hereditarian would suggest that Episcopalianism and Republicanism are directly coded for in the genes.'" Ehrlich and Feldman are debating a strawman opponent. As behavior geneticists have been pointing out for years, familiality of a trait does not necessarily inform about genetic or environmental bases for that trait. The parent-child correlation in normal families is inherently ambiguous, because children in normal families get both genetic and environmental influences from their parents. This is precisely why family studies -- the workhorse of social science-- are uninterpretable, and why genetically informative research designs are needed. No behavior geneticist would suppose that high parent-child correlations in normal families for religious sect membership implies any genetic influence. The "vulgar hereditarian" referred to is a mythological creature, just like the mythical "genetic determinist," who thinks that all behaviors and behavioral variants are "directly coded for in the genes." What's more, religious sect membership has been examined using genetically informative methods. The result of "[t]win studies of religious affiliation (e.g. Christian, Jewish, Muslim) have shown that variance in this trait is nearly completely environmental in origin, thus demonstrating that model-fitting is not intrinsically biased and can indeed show no genetic effects when that is the case" (Bouchard and McGue, 2003, p. 29).



A Gene Shortage?

Ehrlich and Feldman state (p. 92) that "[p]erhaps the most important reason to doubt that genetic variation accounts for a substantial portion of observed differences in human behavior is simply that we lack an extensive enough hereditary apparatus to do the job—that we have a 'gene shortage.'" This bizarre conclusion follows from Ehrlich's mistaken notion that behavior geneticists and evolutionary psychologists assume something like a literal one-to-one correspondence of genes to behaviors, a glaring strawman. In Ehrlich's view, evolutionary psychology requires "genes for" specific behaviors, such as establishing charities, building hydrogen bombs, or publishing books (p. 93). Since this model of gene-behavior relationship has been advocated by absolutely no one, the "gene shortage" problem is no problem at all. As Schoeneman (p. 101) points out, the "suggestion that there aren’t enough genes to code for all possible behavioral responses is irrelevant" because "the argument has never been that there is total genetic control of every aspect of behavior." It would be nevertheless desirable to know how Ehrlich purports to calculate a minimum number of genes required for a given portion of behavioral differences to be accounted for by genetic variation. Obviously this would require an explicit model defining how genes are related to behaviors, but no such model is presented. Would 50 or 100 thousand genes have been enough? Why or why not? Ehrlich and Feldman do not even attempt to make their reasoning explicit.


Genetic Influences on Brain Structure

Ehrlich and Feldman question (p. 93-94) claims for strong genetic influences on individual differences in brain structure. It will be instructive to examine that critique. Ehrlich and Feldman refer to only a single relevant study, the MRI study of Thompson et al (2001), which used a small sample of MZs, DZs, and unrelated control pairs. Ehrlich and Feldman's main criticism of Thompson et al (2001) is that the DZ intrapair correlations for some brain areas are greater than half --but still less than-- the MZ correlation for the same brain areas, which implies significant environmental and genetic influence on brain volumes in those areas (assuming the correlations are not the result of sampling error). Ehrlich and Feldman conclude from this that "[t]he relationship between the MZ and DZ correlations certainly does not suggest 'strong genetic control of brain structure' (p. 1254) or even 'tight coupling of brain structure and genetics' (p. 1256) as claimed."

Ehrlich and Feldman's commentary on Thompson et al (2001) is misleading. First, Ehrlich and Feldman ackowledge (p. 88) that the increase in size in homonid brain size "certainly" occurred in response to natural selection, which of course requires that brain volume in homonids has at least a modest narrow heritability. Second, what Thompson et al (2001) showed is that the volume of some cortical brain regions are strongly genetically influenced, whereas others are less strongly influenced. The degree of genetic control of brain structure was found to be regionally variable, precisely as the authors expected based on prior research (p. 1254). The goal was simply to map this regional genetic variability. Ehrlich and Feldman disagree with Thompson et al's (2001) conclusion that there is 'strong genetic control of brain structure,' but what Thompson et al (2001) actually wrote was that heritability was high "in a large anatomical band spanning frontal (F), sensorimotor (S/M) and Wernicke’s (W) language cortices, suggesting strong genetic control of brain structure in these regions, but not others." By omitting the words 'in these regions, but not others,' Ehrlich and Feldman obscured the obvious answer to why DZ similarities for some brain regions are nearly as great as those of MZs - these are regions for which the volume is relatively weakly influenced by heredity, for instance the perisylvian language and spatial association cortices. Third, even a glance at Thompson et al's (2001) maps (p. 2) makes it very clear that MZs are far more similar than DZs for most regions, and DZs are far more similar than unrelated pairs. For Ehrlich and Feldman to object to claims that brain structure is strongly genetically controlled on the grounds that DZ correlations are high for some regions is not reasonable.


Ehrlich and Feldman's commentary conspicuously omits other relevant data. Thompson et al (2001) utilized a very small sample (n=40; 10 mz pairs, 10 dz pairs), and encouraged caution regarding the interpretation of their results (p. 1254). However, the results of Thompson et al (2001) are strongly supported by a substantial body of additional MRI research, including those using much larger sample sizes. For a review of findings on genetic influences on individual diferences in human brain structure, and a discussion of ongoing genetic brain mapping initiatives, see Thompson et al (2002). None of this research is even alluded to by Ehrlich and Feldman. For instance, Posthuma et al (2000) report a heritability of 0.88 for cerebeller volume, and Posthuma et al (2002) reported a heritability of 0.82 for whole-brain grey-matter volume, and 0.87 for whole-brain white-matter volume, using a much larger sample (n=258). Pennington et al (2000) calculated a heritability of 0.9 for total cerebral volume, based on MRI volume analyses of 34 mz pairs and 32 dz pairs. Baare et al (2001) calculate heritabilities of 0.9, 0.82, and 0.88 for whole brain, grey matter, and white matter volume, respectively, based on an extended twin design including 54 mz pairs, 58 dz pairs, and 34 sibs. Carmelli et al (1999) and Pfefferbaum et al (2000) calculated a heritability of 0.81 for intracranial volume in a sample of 85 elderly male twin pairs. Heritability of brain structure in mice and rhesus monkeys (e.g. weight, number of neurons), which can be reared in identical environments, is also high (Roderick, 1973; Wimer and Wimer, 1989; Strom, 1999; Williams, 2000). I am not aware of an MRI study of brain volume using reared-together adoptive sibling pairs, but a firm prediction is that when they are done, their mean intra-pair correlation will be near 0. In general, individual differences in brain volume are just as heritable as individual differences in height.

Ehrlich and Feldman (p. 93) refer to experience-dependent development of the parts of the human brain, for instance the structure of the visual cortex. For instance, if you rear a cat in a completely dark environment, its visual cortex will not develop properly. However, we should not conclude from these types of experiences that individual differences, even in fine-scale cortical structure, are controlled exclusively by experience. For instance, individual differences in the arrangement of functional cortical columns in the visual cortex, previously thought to result almost entirely from sensory input, have recently been shown to be strongly genetic influenced in cats (Kaschube et al., 2002), and much less dependent upon sensory input than previously thought (Godecke et al., 1997;. Chapman et al., 1999). Kaschube et al. (2002) is one of the first demonstrations of signficant heritability for fine-scale brain structure. This does not show that the brain does not require sensory/environmental inputs to develop, but that the effects of that experience on the developing brain in different individuals are signficantly influenced by genetic factors that differ between individuals.

Of course, this is not to say that the volume of all brain regions, or all aspects of brain structure (e.g. gyral patterns, ventricular volumes), are under equally strong genetic control, and/or cannot be modified significantly by experience. For instance, the volume of the hippocampus, which has a heritability of 0.4 in humans (Sulivan et al., 2001), 0.5 in mice (Lu et al., 2001), and 0.54 in rhesus monkies (Lyons et al., 2001), is clearly affected by experience. For instance, Maguire et al (2000) examined the hippocampi of London taxi drivers (n=16) and matched controls using MRI scans to see if their spatial-memory-demanding job resulted in increases in volume of the hippocampus, which is heavily recruited in spatial navigation and memory. Such differences were found, bilaterally in the posterior hippocampus. The difference was greatest in those who had driven taxis the longest. There were no significant differences in other brain regions. Although the sample size is relatively small, and the effect size is small, and the data is cross-sectional rather than longitudinal, it appears that hippocampal volume increases through extensive spatial-navigation practice.

There are is also a substantial body of research supporting Thompson et al's (2001) finding that brain volume is significantly (~0.2-0.4) correlated with psychometric IQ, which Ehrlich and Feldman mention but do not substantively criticize. This is consistent with the finding that, across primate species, body-size-corrected neocortical volume correlates positively with measures of behavioral complexity (Reader and Laland, 2002). Posthuma et al (2002) found a somewhat smaller correlation of 0.25. Pennington et al (2000) report a correlation of 0.3-0.4 with Wechsler Full-scale IQ. MacLullich et al (2002) found a correlation of 0.39 of intracranial capacity with scores on Raven's Standard Progressive Matrices. Andreasen et al (1993), Raz et al (1993), Wicket et al (1994, 2000), Reiss et al (1996), Tisserand et al (2001), and Allin et al (2001) also report significant correlations of about 0.2-0.4 between measures of brain volume and cognitive ability, while Storfer (1999) and Vernon et al (2000) review research in this area. In the words of Wickett et al., there "is no longer any doubt that a larger brain predicts greater intelligence," and that "[w]hat is required now is a more fine-grained analysis of why it is that a larger brain predicts greater intelligence" (2001, p. 1096). It has also been demonstrated by Wickett et al (2001) that the more "g-loaded" a cognitive test is, the more strongly it correlates with brain volume. Significantly, bivariate genetic analyses suggest that the association of brain volume and psychometric g is mediated by common genetic factors (Posthuma et al., 2002; Thompson et al., 2002). In other words, IQ variance and brain volume variance are apparently pleiotropic effects of the same genes. This is consistent with a a simple model in which genes influencing brain volume, which in turn influences cognitive ability, or a model in which both variables being influenced by a third factor such that the correlation of brain volume and IQ is noncausal. Finally, PET imaging of subjects performing 'g'-loaded spatial and verbal tests shows that such tasks differentially recruit lateral frontal cortical areas (Duncan et al., 2000), the very brain region whose volume is under the strongest genetic control.



References

Allin et al., 2001. Cognitive and motor function and the size of the cerebellum in adolescents born very pre-term. Brain 124, pp. 60-66.

Baare et al., 2001. Quantitative genetic modeling of variation in human brain morphology. Cerebral Cortex 11, pp. 816-824.

Bartels et al., 2002. Genetic and environmental influences on the development of intelligence. Behavior Genetics 32, pp. 237-249.

Bartley et al., 1997. Genetic variability of human brain size and cortical gyral patterns. Brain 120, pp. 257-269.

Bouchard, T.J., and McGue, M., 2003. Genetic and envionmental influences on human psychological differences. Journal of Neurobiology 54, pp. 4-45.

Cadoret et al., 1983. Evidence for gene-environment interaction in the development of adolescent antisocial behavior. Behavior Genetics 13, pp. 301-310.

Cadoret et al., 1995. Genetic-envionmental interaction in the genesis of agressivity and conduct disorders. Archives General Psychiatry 52, pp. 916-924.

Caspi et al., 2002. The role of genotype in the cycle of violence in maltreated children. Science 297, pp. 851-854.

Chapman et al., 1999. Development of orientation preference in the mammalian visual cortex. Journal Neurobiology 41, pp. 18-24.

Devlin et al., 1997. The heritability of IQ. Nature 388, pp. 468-471.

Dobzhansky, T., 1964. Heredity and the nature of man. New York: Harcourt, Brace, and World.

Drayna et al., 2001. Genetic correlates of musical pitch recognition in humans. Science 291, pp. 1969-1972.

Duncan et al., 2000. A neural basis for general intelligence. Science 289, pp. 457-460.

Ehrlich, P., and Feldman, M., 2003. Genes and culture: What creates our behavioral phenomes? Current Anthropology 44, pp. 87-107.

Geschwind et al., 2002. Heritability of lobar brain volumes in twins supports genetic models of cerebral laterality and handedness. Proceedings of the National Academy of Sciences 99, pp. 3176-3181.

Godecke et al., 1997. Development of orientation preference maps in area 18 of kitten visual cortex. European Journal Neuroscience 8, pp. 1754-1762.

Jacobs et al., 2001. Heritability Estimates of Intelligence in Twins: Effect of Chorion Type. Behavior Genetics 3, pp. 209-217.

Kaschube et al., 2002. Genetic influence on quantitative features of neocortical architercture. Journal of Neuroscience 22, pp. 7206–7217.

Lu et al., 2001. Complex trait analysis of the hippocampus: mapping and biometric analysis of two novel gene loci with specific effects on hippocampal structure in mice. Journal of Neuroscience 15, pp. 3503-3514.
Lyons et al., 2001. Early life stress and inherited variation in monkey hippocampal volume. Archives General Psychiatry 58, pp. 1145-1151.

Maguire et al., 2000. Navigation-related structural change in the hippocampi of taxi drivers. Proceedings of the National Academy of Sciences 97, pp. 4398-4403.

Mayeux et al., 1995. Synergistic effects of traumatic head injury and apolipoprotein-e4 in patients with Alzheimer's disease. Neurology 45, pp. 555-557.

McCall, R., 1991. So many interactions, so little evidence. Why? In Wachs. T.D., and Plomin, R., (eds.), Conceptualization and measurement of organism-environment interaction. Washington, DC: American Psychological Association, pp. 146-161.

McGue, M., 1997. The democracy of the genes. Nature 388, pp. 417-418.

Niesser et al, 1996. Intelligence: knowns and unknowns. Report of a task force established by the board of scientific affairs of the American Psychological Association. American Psychologist 51, pp. 77-101.
Otto, S.P., 2001. Intelligence, Genetics of: Heritability and Causation, in International Encyclopaedia of the Social and Behavoural Sciences, Smelser, N.J., and Baltes, P.B., eds. Elsevier. pp.7651--7658

Pennington et al., 2000. A twin MRI study of size variations in human brain. Journal of Cognitive Neuroscience 12, pp. 223-232.

Pfefferbaum et al., 2000. Brain structure in men remains highly heritable in the seventh and eighth decades of life. Neurobiology of Aging 21, pp.63-74.

Posthuma et al., 2000. Multivariate genetic analysis of brain structure in an extended twin design. Behavior Genetics 30, pp. 311-319.

Posthuma et al., 2002. The association between brain volume and intelligence is of genetic origin. Nature Neuroscience 5, pp. 83-84.

Reader, S.M., and Laland, K.N., 2002. Social intelligence, innovation, and enhanced brain size in primates. Proceedings of the National Academy of Sciences 99, pp. 4436-4441.

Reiss et al., 1996. Brain development, gender and IQ in children. A volumetric imaging study. Brain 119, pp. 1763-1774.

Roderick et al., 1973. Genetic and phenotypic variation in weight of brain and spinal cord between inbred strains of mice. Brain Research 64, pp. 345–53.

Rowe, D.C., and Jacobson, K.C., 1999. In the mainstream: Research in behavioral genetics. In : Carson and Rothstein, eds. Behavioral Genetics: The Clash of Culture and Biology. Baltimore, London: Johns Hopkins University Press, pp. 12-34.

Rutter and Silberg, 2002. Gene-environment interplay in relation to emotional and behavioral disturbance. Annual Reviews Psychology 53, pp. 463-490.

Strom, R.C., 1999. Genetic Analysis of Variation in Neuron Number. Dissertation, Univ Tenn, Memphis.

Storfer, M., 1999. Myopia, intelligence, and the expanding human neocortex: Behavioral influences and evolutionary implications. International Journal of Neuroscience 98, pp. 153–276.

Sullivan et al., 2001. Heritability of hippocampal size in elderly twin men: equivalent influences from genes and environment. Hippocampus 11, pp. 754-762.

Thompson et al., 2001. Genetic Influences on Brain Structure. Nature Neuroscience 4, pp. 1253-1258.

Thompson et al., 2002. Mapping Genetic Influences on Human Brain Structure. Annals of Medicine 34, pp. 523-536.

Tisserand et al., 2001. Head size and cogntive ability in nondemented older adults are related. Neurology 56, pp. 969-971.

Tramo et al., 1998. Brain size, head size, and intelligence quotient in monozygotic twins. Neurology 50, pp. 1246-1252.

Wickett et al., 2000. Relationships between factors of intelligence and brain volume. Personality and Individual Differecences 29. pp. 1095-1122.

Williams, R.W., 2000. Mapping genes that modulate brain development: a quantitative genetic approach. In: Mouse brain development (Goffinet AF, Rakic P, eds). Springer Verlag, New York, pp 21–49.

Wimer, C. C., and Wimer, R. E., 1989. On the sources of strain and sex differences in granule cell number in the dentate area of house mice. Developmental Brain Research 48, pp. 167–176.

Wright et al., 2002. Genetic contributions to regional variability in human brain structure: methods and preliminary results. Neuroimage 17, pp. 256-271.

Vernon et al., 2000. The neuropsychology and psychophysiology of human intelligence. In R. J. Sternberg (ed.), Handbook of intelligence (pp. 245-264). Cambridge: Cambridge University.