The cognitive ability of blacks raised by non-blacks

Last Updated on April 30, 2022

In a previous post, I argued that the black-white cognitive ability gap is responsible for many of the undesirable social disparities that we find between blacks and whites in the United States – including disparities in income, education, occupation, and crime. This means that the cause of the cognitive ability gap is extremely important, as this informs how (and if) we can reduce these social disparities. In another post, I concluded that many common environmental explanations of the gap – test bias, schooling, socioeconomic status, and societal racism – all failed to account for the gap. I also examined some arguments for a genetic explanation of the gap in another post, where I concluded that we should use direct data to quantify the degree to which the black-white cognitive ability gap is due to genetic differences vs environmental differences. One form of direct data is data on the cognitive ability of blacks raised by non-blacks. In this post, I will attempt to review all such data. I conclude that, while much of the data is low quality and no conclusive judgments can be made, most of the data is compatible with a primarily environmental explanation of the gap.

Notes on method

Direct data

As I explained in another post, we should use direct data to quantify the degree to which racial differences in cognitive ability are due to genetic differences vs environmental differences. By “direct” data, I mean data that measures the cognitive ability of random samples of black children raised in sufficiently enriched environments. By “sufficiently enriched” environments, I mean environments that eliminate the relevant environmental differences between black and white children. As I argued in the other post, this kind of data best isolates the relative effect of environmental vs genetic influences on the cognitive ability. Mackintosh (2011) has also written that the “critical experiment” for determining whether genetics are responsible for the gap involves a method where we “take a random sample of black and white children at birth and bring them up in carefully matched adoptive homes or other comparable environments and measure their IQ scores at age 10 or so” (page 150).

One example of direct data is data from transracial adoption studies, e.g., adoption studies that involve black children adopted by white parents. Rushton and Jensen (2005) have asserted that “Transracial adoption studies provide one of the best methods for resolving the question of group differences in mean IQ” (page 275). These studies (ideally) eliminate most of the relevant environmental differences between blacks and whites (e.g. differences in SES, parenting practices, etc.), which means the black children are raised in sufficiently enriched environments. This enables us to quantify the degree to which the cognitive ability gap is due to genetic differences vs environmental differences. For example, if we perform an ideally perfect transracial adoption study (i.e., a random sample of black children are randomly assigned to white families at birth), then any cognitive ability gap that obtains between black children and white children raised by white families can be attributed to genetic differences. Of course, even an ideally perfect transracial adoption study cannot control for all environmental factors (e.g., prenatal environmental factors, possible societal racism, etc.). Nevertheless, I believe this is the best way to control for the isolated effects of genetic and environmental factors on the cognitive ability gap.

In chapter “Group Differences in intelligence” of the book Handbook of intelligence, Loehlin (2000) also remarked (page 185) on the usefulness of transracial adoption studies to settle whether genetics are responsible for the cognitive ability gap:

When infants of one racial group are reared by parents of another, if the children tend to display the characteristics of the adopting group, it is prima facie evidence for postnatal environmental effects, and if they tend to display the characteristics of the group from which they came, it is prima facie evidence for the genes or prenatal effects.

Transracial adoption studies are just one source of direct data. Other studies involve black and white children raised by similar caretakers (e.g., children raised in nurseries) or black/biracial children raised by white mothers/parents. All of these studies provide direct data because they involve black children raised in sufficiently enriched environments. I do not consider data involving black children raised by high-SES black parents as “direct” data because such data does not control for environmental factors that may correlate with having a black parent (e.g., parenting differences). In this post, I will consider all instances of direct data that I have come across. Unfortunately, much of the data is low quality, but evaluating the data can still serve as useful hints to know what is suggested by the available research.

Three models

In their paper “Thirty years of research on race differences in cognitive ability”, Rushton and Jensen (2005) [archived] evaluated thirty years of findings relevant to the debate on whether genetic differences are responsible for racial differences in cognitive ability. They framed the debate by presenting two models: the hereditarian model posits that any portion (say, 50%) of the cognitive ability gap is due to genetic differences, and the culture-only model posits that 0% of the cognitive ability gap is due to genetic differences (pages 237-238). For each finding, they assessed the degree to which the finding provided evidential support to each model. They concluded that the hereditarian model is best supported by the preponderance of evidence.

I approve of this general framework for assessing competing models to the debate, i.e. reviewing how well each relevant finding supports each model. I plan to perform a similar analysis in my post. However, my analysis differs from Rushton and Jensen’s analysis in two ways. Firstly, I will only consider direct data (as I defined above) whereas Rushton and Jensen considered a wider range of data.

Secondly, I object to framing the debate using the hereditarian and culture-only models so described. This framing gives the impression that the important question in the debate is whether any portion of the cognitive ability gap is due to genetic differences. In fact, Rushton and Jensen note that “the defining difference [between the hereditarian and culture-only models] is whether any significant part of the mean Black–White IQ difference is genetic rather than purely cultural or environmental in origin” (page 238). I do not believe that this is the important disagreement. Now, some (maybe even many) environmentalists argue that genetic differences play no role in the cognitive ability gap (e.g., Nisbett 2005), but I believe these environmentalists are mistaken to argue for such a strong position. I agree with Hunt (2011) who has written (page 434) that what is important is not the existence of the gap, but rather the magnitude of the gap (if it exists):

the 100% environmental hypothesis is something of a stalking horse. Many researchers who are primarily interested in environmental differences associated with racial and ethnic differences in intelligence would not be at all perturbed by an ironclad demonstration that, say, 3% of the gap is due to genetic differences. The real issue is over the identity and size of genetic and environmental influences on group differences in intelligence, not the existence of either one.

So the important question is not whether any portion the cognitive ability gap is due to genetic differences. The important question is to what degree is the cognitive ability gap due to genetic differences. I believe we can split possible answers to this question into three broad models:

The environmental model: Genetic differences are responsible for minor black-white gaps, i.e. genetic differences are responsible for a gap of less than one-third of a standard deviation. For example, this model posits that less than 5 points of the IQ gap is due to genetic differences.
The genetic model: Genetic differences are responsible for major black-white gaps, i.e. genetic differences are responsible for a gap of more than two-thirds of a standard deviation. For example, this model posits that more than 10 points of the IQ gap is due to genetic differences.
The intermediate model: Genetic differences are responsible for moderate black-white gaps, i.e. genetic differences are responsible for a gap of between one-third and two-thirds of a standard deviation. For example, this model posits that between 5 and 10 points of the IQ gap is due to genetic differences.

I choose these models for a few reasons.

Firstly, the black-white cognitive ability gap is typically measured at around one standard deviation across a variety of tests (IQ scores, educational tests, industrial tests, etc.) and contexts (Roth et al. 2001). Therefore, these three models map on nicely to the possible percentages of the gap that might be attributed to genetic differences. If you grouped the possible percentages of the gap attributed to genetics into three equal ranges, the ranges would be: (a) genetics explain 0-33% of the gap, (b) genetics explain 33-67% of the gap, and (c) genetics explain 67-100% of the gap. These three ranges correspond to the three models given above. Now, for some tests of cognitive ability, the black-white gaps are not exactly one standard deviation, but they are typically close enough that this point still holds.

Secondly, unlike Rushton and Jensen’s two-model framing, my framing captures what is actually practically important in the race and cognitive ability debate. For example, it does not make an important practical difference whether genetic differences account for 0 points or 2 points of the IQ gap. On my framing, the same model (the environmental model) would be true under both possibilities, indicating (correctly) that there’s no important difference between the two possibilities. On Rushton and Jensen’s framing, the culture-only model would be true if the gap were 0 points and the hereditarian model would be true if the gap were 2 points, suggesting (incorrectly) that there’s an important difference between these two possibilities. For another example, it makes a very large practical difference whether genetic differences are responsible for 2 points or 12 points of the IQ gap. On my framing, one model would be true if the gap were 2 points (the environmental model) and another model would be true if the gap were 12 points (the genetic model), indicating (correctly) that there is an important difference between these two possibilities. On Rushton and Jensen’s framing, the same model (the hereditarian model) would be true under both possibilities, suggesting (incorrectly) that there’s no important difference between these two possibilities.

My post will assess which of these three models is supported most by each of the findings that I review below. My goal is not to assess whether any portion of the gap is caused by genetic differences. Rather, my goal is to (roughly) quantify the portion of the gap that is caused by genetic differences. Note that I attempt to quantify the portion of the gap due to genetic differences by using these three rough models rather than by trying to calculate the precise percentage of the gap due to genetic differences. This is because there isn’t enough data to give a fine-grained calculation with a high degree of confidence. So I am forced to (try to) quantify the genetic component of the gap in broad terms.

Assessing the models

That said, this post will involve reviewing each piece of direct data that I have come across. For each piece of data, I will highlight the main findings that require explanations. For each finding, I will assess which model is provided with the most evidential support by the finding. This involves determining whether the finding would be predicted and explained by each model. If a finding would be explained or predicted by one model but not another, then the finding provides more evidential support for the former model. If a finding would be explained or predicted by multiple models, then the model that can provide the more parsimonious explanation (i.e., can explain the findings with the fewest assumptions) will be favored. Also, if a finding directly contradicts a model, then that provides strong evidence against the model.

Also, for each piece of data considered, I will note Rushton and Jensen (2005)‘s assessment of the same data.

Studies

A note on language before proceeding: unless the context indicates otherwise, “black” describes a person two black parents and “biracial” describes a person with one black and one white parent.

Black and white children adopted by white families

The first study that I will consider – the Minnesota Transracial Adoption Study (MTAS) [archived] – is commonly invoked by hereditarians in support of their theory. This study examined the IQs of 130 black and black-biracial children and the IQs of 25 white children who were adopted by middle-class white families. The study involves two waves of results. In wave 1, Scarr and Weinberg (1976) [archived], measured the IQs of the adoptees at age 7. The IQs for the black (n=29), biracial (n=68), and white adoptees (n=25) were 96.8, 109.0, and 111.5, respectively (Table 2). In wave 2, Scarr and Weinberg (1992) [archived] retested the IQs of the children at age 17 (although some of the participants had dropped out of the study). The mean IQ scores for all racial groups diminished. The respective IQs for black (n=21), biracial (n=55), and white (n=16) adoptees were 89.4, 98.5, and 105.6 (Table 2). “Black” in this context refers to adoptees with two black biological parents. The IQs of the black adoptees at age 17 was not much larger than the IQs of their non-adopted racial peers. If this study did indeed equalize the environments for blacks and whites, then the results would seem to indicate that the racial IQ gap is largely caused by genetic differences. Indeed, Rushton and Jensen (2005) spend several pages describing this study as evidence of their hereditarian model (pages 256-259). However, there are a few reasons to caution against this interpretation:

Attrition favored the white children more than the black/biracial children. Many of the low-scoring White adoptees from the original measurement did not return for the follow-up study. We know this because the age-7 average IQ of the original White group was 111.5, yet the age-7 average IQ of the White group that appeared for follow-up testing was 117.6, over 6 points greater than the original group (Table 2). For both the black/biracial children, there was no significant IQ difference between the age-7 IQs of the original group and the follow-up group . Differences in attribution does not explain the IQ gap at age 7, but it does explain some of the gap at age 17. Thomas (2017) [archived] found that correcting for attrition resulted in adjusted age-17 IQs of black, biracial, and white children of 90.1, 98.3, and 101.8, respectively (Table 2). This reduces the black-white gap from 16.2 to 11.7 points and the biracial-white gap from 7.1 to 3.5 points.
The black children were adopted substantially later than the other children. The average white adoptee was adopted at 19 months old (Table 5), but the average black child was adopted at 32 months old (Table 10). The study showed that late adoptees had much lower IQs than early adoptees, particularly for the black children. Black/biracial children adopted within one year of birth had a mean IQ of 99, whereas later adopted black/biracial children had a mean IQ of 92 (See Table 2 of the follow-up study). Unfortunately, there is not enough information to determine if the late black/biracial adoptees scored lower because they were disproportionately black (as opposed to biracial) or if the black adoptees scored lower because they were disproportionately adopted late, so we cannot confirm if age of adoption had an independent effect. Nevertheless, this is an important factor to keep in mind. Rushton and Jensen (2005) have argued that age of adoption “does not influence children’s IQ scores after age 7” (page 259), but this has been challenged by Thomas (2017).
Even though the study may have equalized environments after adoption, it could not have equalized environments before adoption. For example, a high blood lead level in infants can result in a noticeable reduction in IQ during adulthood. In fact, a 10 μg/dL increase in blood lead level for infants aged 24-months was associated [archived] with a 5.8-point decline in their age-10 IQ (page 857). Blood lead level might have been particularly important for the subjects of the MTAS because, when it was originally published in the 1970s, black children aged 6 months-5 years had much higher blood lead levels [archived] (page 6) than similarly aged white children: over half of Black children (52%) in this age range had a blood lead level greater than 20 μg/dL, compared to only 18% of white children. Blood lead levels is just one example. The point is that racial differences in pre-adoptive environments (whether prenatal or postnatal) could have contributed to the IQ gap. This may have been particularly important for the black adoptees given their late age of adoption (32 months). Of course, we have no direct evidence that the black adoptees were subject to inferior pre-adoptive environments, but it is a factor to keep in mind.

That being said, there are two main sets of findings that require explanation regarding the black-white IQ gap (I ignore the biracial-white IQ gap because the the size of the gap was only 2.5 ± 3.5 points, which is small enough to be “statistical noise” as Thomas (2017) has pointed out):

Firstly, there was a significant black-white IQ gap at age 7. How well are these findings explained by the genetic and environmental models?

Environmental model: the environmental model can explain these findings only if we posit certain unverified assumptions, e.g., that the black children suffered inferior pre-adoptive environments. I do not consider this to be an ad hoc assumption because there is some independent evidence in support of this assumption (see points 2 and 3). Nevertheless, we do not have direct evidence to verify this assumption, so they are unverified assumptions.
Genetic model: The genetic model predicts these findings (without requiring any unverified assumptions). The genetic model predicts that at least 10 points of the IQ gap is due to genetic differences, which is consistent with the finding here which finds an IQ gap of about 14.9 points between blacks and whites.

Secondly, the magnitude of the black-white IQ gap did not grow between ages 7 and age 17 (after correcting for attrition). How well are these findings explained by the genetic and environmental models?

Environmental model: the environmental model is consistent with these findings, but it neither predicts nor explains these findings. The environmental model makes no claims either way on how the black-white IQ gap among children raised in similar environments fluctuates over time.
Genetic model: the genetic model is contradicted by these findings. The genetic model predicts that the IQ gap would widen as children age (assuming similar environments over time) because the heritability of IQ increases dramatically as children age (see the following reviews of intelligence research: Neisser et al. 1996 report that the heritability of IQ increases from 0.45 in childhood to 0.75 in late adolescence. Plomin et al. 2017 report a heritability of around 41% at age 9 to 66% at age 17. Plomin and Deary 2015 [archived] report that “heritability increases linearly, from (approximately) 20% in infancy to 40% in adolescence, and to 60% in adulthood”. Bouchard 2013 [archived] reviews key twin and adoption studies to find “heritability increases with age until late adulthood”). Even Rushton and Jensen (2005) have even asserted that “Trait differences not apparent early in life begin to appear at puberty and are completely apparent by age 17” (page 259). They used this point to discount transracial adoption studies that did not find racial gaps for young children. Therefore, the genetic model is contradicted by these findings, unless one posits certain ad hoc assumptions (e.g., the environment for the black children improved over time to offset the increasing influence of their disadvantaged genes).

The genetic model better explains the first set of findings whereas the environmental model better explains the second set of findings. Which models fares better overall? It seems to me that the environmental model fares better because the assumptions needed for the environmental model to explain the findings at age 7 (e.g., that the black adoptees were exposed to inferior pre-adoptive environments) are more reasonable than the assumptions needed for the genetic model to explain why the gap did not widen (e.g., that the environment for the black adoptees improved over time relative to the environment for white children, offsetting the increasing influence of their disadvantaged genes). That said, I will be generous to the genetic model and conclude that the two models fare equally well. That is, I will conclude that, regarding the IQ gap, the results of this study are ambiguous with respect to whether they provide more evidential support for the genetic or environmental model. In fact, Scarr (1998), one of the original authors, states that the results of the study “can be used to support either a genetic difference hypothesis or an environmental difference one” (page 230).

It should be noted that, despite the small IQ gains, the black and biracial children saw much higher levels of academic achievement than their non-adopted racial peers.

The white adoptees scored in the 62nd, 58th, and 56th percentiles in vocabulary, reading, and mathematics scores, respectively (see Table 6 of the follow-up study).
The biracial adoptees scored in the 60th, 59th, and 50th percentiles in those respective areas.
The black adoptees scored in the 54th, 48th, and 36th percentiles in those respective areas.

If we assume the achievement scores are normally distributed, we can calculate these racial gaps in standard deviations (SDs) using a z-score table. The white adoptees scored +0.31, +0.20, and +0.15 SDs from the mean in vocabulary, reading, and mathematics scores, respectively. The biracial adoptees scored +0.25, +0.23, +0.00 SDs from the mean in those respective areas. The black adoptees scored +0.10, -0.05, and -0.36 SDs from the mean in those respective areas. Thus, the biracial-white gap (white score minus biracial score) in vocabulary, reading, and mathematics scores were about 0.06, -0.03, and 0.15 SDs. The black-white gap (white score minus black score) in those respective areas were about 0.21, 0.25, and 0.51 SDs.

Note the achievement scores of the black adoptees are much higher than the scores for black students in the general population (see my previous post here). Also note that the comparisons with the white children are misleading because they were not adjusted for attrition (which favored the white children). Adjusting for attrition would likely reduce achievement gaps even further. Regardless, even without adjusting for attrition, the study provides evidential support for the environmental model as an explanation of the vocabulary/reading achievement gap (since the gaps between blacks and whites are less than one-third of a deviation). Furthermore, the study provides evidential support for the intermediate model as an explanation of the mathematics achievement gap (since the gap between blacks and whites is between one-third and two-thirds of a standard deviation). After adjusting for attrition, it is very plausible that the mathematics achievement gap would reduce to less than 0.33 SDs, thus providing support for the environment model as an explanation of the mathematics achievement gap as well.

Rushton and Jensen (2005) spend several pages reviewing and analyzing the results of this study (pages 256-259). They argue that the data supports a hereditarian interpretation (page 259):

The mean IQ and school achievement scores of Black children reflected their degree of African ancestry. At both age 7 and 17, the adopted children with two Black biological parents had lower average IQs and school achievement scores than did those with one Black and one White biological parent, and these children, in turn, averaged lower scores than did those with two White biological parents.

They are correct that the racial gaps in test scores is evidence of the hereditarian model, because the hereditarian model would predict such gaps. However, they neglect to mention the fact that the gaps did not widen between ages 7 and 17, a fact that contradicts the hereditarian model.

Biracial and white children raised in Germany

Eyferth (1959) [archived] studied the IQs of 83 white and 98 biracial (black-white) children in Germany who were born between 1945 and 1953. The mothers of the children were all white Germans, mostly of low socioeconomic status. Their fathers were either white or black members of the US occupation forces. The white children had a mean IQ of 97.2, whereas the biracial children had a mean of 96.5, a negligible difference. The children were between the ages of 5 and 13 (average age of 10) at the time of testing. The environmental model is the most straightforward explanation of the data found here. However, I should mention that there are some issues that limit the strength of this interpretation.

Firstly, the black and white samples were not representative of the average population, because about 30% of black applicants were rejected admission to the armed forces based on qualification exams, while only about 3% of whites were rejected. Thus, one could argue that this was a genetically elite sample of blacks, and so the parity achieved here cannot be extrapolated to the general black population. However, if there is a significant genetic explanation of the racial IQ differences, this would predict that the white children in the study would have significantly higher scores than the biracial children because:

The white soldiers had higher IQs than the black soldiers. Even though the admission test may have filtered out candidates below a certain cognitive threshold, the IQ scores for the white soldiers should still be distributed higher than that of the black soldiers (e.g. imagine removing all men and all women below 5’2 from the population; the resulting population would still have men taller than women, on average). In fact, Jencks and Phillips (1998) [archived] noted that James Flynn’s analysis of the data suggests that the Black-White gap for the soldiers was about four-fifths that of the general population (see footnote 47 on page 19). Thus, despite the disproportionate acceptance rates, there was still a ~12 point IQ gap between the black and white soldiers. Insofar as the BW IQ gap is genetic, similar differences should be inherited in the children.
Regression to the mean. Even if though the black soldiers in the study were a slightly elite sample, the higher IQs of the black soldiers could not have been inherited completely by the black children because, assuming that the BW IQ gap is mainly genetic, the black children IQs would regress to the genetic black mean, while the white children IQs would regress to the genetic white mean. For example, Rushton and Jensen (2005) [archived] (section 9) argue that this regression should occur insofar are racial IQ differences are genetic (for context, Rushton and Jensen were perhaps the most prominent academic advocates of genetic causes of black-white IQ differences):

For any trait, scores should move toward the average for that population. So in the United States, genetic theory predicts that the children of Black parents of IQ 115 will regress toward the Black IQ average of 85, whereas children of White parents of IQ 115 will regress toward the White IQ average of 100.

So if racial IQ differences are mainly genetic, then regression to the mean should occur. This means that even though the IQ gap of the soldiers was ~3 points smaller than the gap in the general population (12 vs 15), much of this gap reduction would not have applied to the IQs of the children. Note that regression to the mean is found in other genetic traits such as height. This regression is quantified by geneticists by an equation known as “breeder’s equation”.

Secondly, another issue with the study is that about 20-25% of the black group was actually French North African. However, if the genetically black soldiers, who constituted 75-80% of the group, had significantly lower genotypic IQs, this would be enough to cause a significant IQ gap between the biracial and white children. The only way this wouldn’t be true is if the French North African soldiers had extremely large IQs to offset the lower IQs of the biracial children, but there’s no reason to assume this.

Finally, one remaining issue with the study is that there may have been some error in the sampling. The mean IQ for the white girls in the study was about 8 points lower than the mean IQ for the white boys in the study. This gap is extremely greater than the gap typically found between boys and girls, which is usually either null or only slightly in favor of males. Furthermore, the study found only a miniscule 1-point IQ gap between black girls and black boys. These points raise the possibility of sampling error in the study. This is another reason to refrain from drawing strong conclusions from this study.

That being said, how well are these findings explained by the genetic and environmental models?

Environmental model: the environmental model predicts these findings. The environmental model predicts that there will not be a large IQ gap between white children and biracial children when raised in similar environments.
Genetic model: these findings contradict the genetic model because the model predicts that white children will have higher IQs than black-white biracial children when raised in similar environments. The genetic model is contradicted unless one posits certain ad hoc assumptions (e.g., the biracial children were given superior environments compared to the white children, the biracial children had higher-IQ mothers, etc. which compensated for the negative effects of their African ancestry).

For these reasons, I conclude that this study provides weak evidential support for the environmental model of the IQ gap.

In their section on racial admixture, Rushton and Jensen (2005) state that the results of this study are “ambiguous” because of the young age of the children, a portion of the black group were French North Africans, and a larger share of the black applicants were rejected entry (page 261). I have already addressed the later two points. I address the issue of young age of testing later in the post.

Black-Japanese and white-Japanese children raised in Japan

Kirkegaard, Lasker, and Kura (2019) [archived] replicated the results of the Eyferth study. Kirkegaard and colleagues investigated the IQs of a number of biracial children in a foster home. The fathers of these children were U.S. servicemen after World War 2 and the mothers were all indigenous Asian women in northeast Asian countries. Two IQ tests were administered to different sets of children: in 1950, IQ tests were given to 22 white-Japanese and 6 black-Japanese children aged 3-4 years old; in 1967, IQ tests were given to 28 white-Japanese and 20 black-Japanese children aged 6-16 years old. The results of both tests reveal a tiny IQ gap (less than one IQ point) between the Black-Japanese and white-Japanese children.

Like the Eyferth study, the clearest and most parsimonious interpretation of these findings is that, counter the genetic model, higher degrees of African ancestry is not associated with inferior genes related to cognitive ability. However, also like the Eyferth study, the study suffers from some problems that limit this interpretation. First, the study may involve children fathered from an unrepresentative sample of black males (e.g., because, relative to whites, a larger proportion of black applicants were rejected admission to the armed forces due to qualification exams). I addressed this issue already with the Eyferth study. Another issue is that the children were “living in foster homes mainly for abandoned children”, which suggests that the children may have had low-IQ mothers (along with whatever environmental or genetic effects that might have). This could possibly reduce cognitive differences between the white and black biracials. Regardless, if the genetic model were true, there should still be some additional disadvantage associated with having higher degrees of African ancestry.

That being said, how well are these findings explained by the genetic and environmental models?

Environmental model: the environmental model predicts these findings. The environmental model predicts that there will not be a large IQ gap between children with different amounts of African vs European ancestors when raised in similar environments.
Genetic model: these findings contradict the genetic model because the model predicts that children with higher degrees of African ancestry will be less intelligent than children with higher degrees of European ancestry when raised in the same environment. The genetic model is contradicted unless one posits certain ad hoc assumptions that cannot be verified (e.g., the black biracial children were given superior environment, the black biracial children had higher-IQ mothers, etc. which compensated for the negative effects of their African ancestry).

For these reasons, I conclude that this study provides weak evidential support for the environmental model of the IQ gap. I believe the study provides only weak evidential support because it’s plausible that the children from this study are an unrepresentative selection of children, since the children were in foster homes for abandoned children. It’s possible that this sample disproportionately selected for children with disadvantaged environments and/or genes (genes because parents who abandon their children are probably likely to have lower IQs themselves), which may mask racial differences in cognitive ability.

Black and white children raised in British nurseries

Tizard (1972) investigated 85 black (West Indian), white, and biracial children aged 2-5 years raised in British long-stay residential nurseries. The children were given psychological tests to determine their cognitive abilities. The scores were normalized to give a mean of 100 and a standard deviation of 10. On the Reynell Comprehension test, the white children (n=39) scored 102.6 and the black/biracial children (n=46) scored 106.3 (page 351). On the Reynell Expression test, the white children scored 98.5 and the black/biracial children scored 98.6. On the Minnesota Nonverbal test, the white children scored 101.3 and the black/biracial children scored 107.7. Averaging the results of this study reveals an IQ gap of about 3.4 points (about one third of a standard deviation) in favor of the black/biracial children. Apparently there is a fourth test, but I didn’t manage to see it in the pdf.

How well are these findings explained by the genetic and environmental models? Neither model predicts these findings. Both models have to posit that the black/biracial children were exposed to superior environments and/or that the black/biracial were an unrepresentative genetically elite sample. Since both models need to posit the same assumptions to explain the findings, this study does not provide strong evidential support for either model. However, the study does provide slightly more evidential support for the environmental model. The reason is that the genetic model would have to posit less likely assumptions than the environmental model in order to explain these findings. For example, if the genetic model were true, the black/biracial children would need a greater environmental advantage to offset this genetic disadvantage. Additionally, if the genetic model were true, it would be less likely that a sample of black/biracial children would be genetically advantaged relative to a sample of white children, Therefore, I conclude that this study provides weak evidential support for the environmental model of the IQ gap.

In their section on transracial adoption studies, Rushton and Jensen (2005) quickly introduce and dismiss this study (and the Moore study below) within one paragraph while noting “to be more informative, future studies need to be supplemented by follow-up testing, as in the Minnesota Study” (page 259). This comes after devoting several pages of discussion to the Minnesota Transracial Adoption study (pages 256-259). I address the issue of young age of testing later in the post.

Biracial children with a white versus black mother

Nisbett (2005) [archived] argued, “If the Black–White IQ gap is largely hereditary, then children having one Black and one White parent should have the same IQ on average, regardless of which parent is Black” (page 305). However, among black-white biracial children, it does matter which parent is black. He reviews Willerman et al (1974) [archived] which finds that among 129 biracial (black-white) 4-year-olds, biracial children with white mothers had a mean IQ that was 9 points higher than biracial children with black mothers (102 vs 93, Table V).

Interestingly, the IQ gap between biracial children raised by black vs white mothers does not seem to be explained by SES differences. Compared to the black-mother/white-father pairs, the white-mother/black-father pairs had an SES index that was only 0.1 SDs higher, they had lower incomes, they had only 0.84 more years of maternal education, and only 0.87 more years of paternal education (page 86). Furthermore, the IQ gap persists even after controlling for marital status. Among married mothers, biracial children with white mothers had a mean IQ of 105 whereas those with black mothers had a mean IQ of 96 (IQ gap = 9 points). Among single mothers, biracial children with white mothers had a mean IQ of 99 whereas those with black mothers had a mean IQ of 88 (IQ gap = 11 points).

The most parsimonious explanation of these findings is that the environment associated with the white mothers is superior to the environment associated with the black mothers (e.g., differences in parenting practices, prenatal differences, etc.). How well are these findings explained by the genetic and environmental models?

Environmental model: the environmental model predicts these findings. The environmental differences that explain the black-white IQ gap in the general population are also present between black and white mothers in this study. Plausibly, the environmental differences would be even larger if we compared black-mother/black-father couples with white-mother/white-father couples (as opposed to comparing black-mother/white-father couples with white-mother/black-father couples), consistent with the environmental model.
Genetic model: these findings do not contradict the genetic model, but the genetic model has difficulty explaining these findings. As stated earlier, the most parsimonious explanation of these findings is that the environment associated with the white mothers is superior to the environment associated with the black mothers. The genetic model cannot explain why similar environmental differences are not responsible for a substantial portion of the black-white IQ gap within the general population.

In summary, this study provides strong evidential support for the environmental model.

Rushton and Jensen (2005) suggest that the findings from this study may not be “interpretable” (page 262):

Willerman, Naylor, and Myrianthopoulos (1974), assuming White mothers provide better pre- or postnatal environments for their children than do Black mothers, interpreted their data as more consistent with a cultural than a genetic hypothesis (see also Nisbett, 1998). However, Loehlin et al. (1975, p. 126) noted that the mixed-race pairs with White mothers averaged almost a year more schooling than did the pairs with Black mothers. Thus the White mothers may have had a higher average IQ than the Black ones. The mid-parent IQs have to be the same for the results to be interpretable.

It is true that the mid-parent IQs must be equal for the results to be interpretable. However, there is no independent, non ad hoc reason to believe that the white-mother/black-father pairs had substantially higher mean IQs than the black-mother/white-father pairs, at least not to such a degree that they would transmit a 9 point IQ advantage to their children. No, one year more of education is not associated with a 9 point IQ advantage (if anything, the absence of large SES differences between the two groups suggests no large IQ differences, since IQ correlates highly with SES attainment). Therefore, the idea that the white-mother/black-father pairs were an unrepresentative genetically elite sample compared to the black-mother/white-father pairs is clearly an ad hoc assumption needed to rescue the genetic model.

Black children adopted by white families versus black families

Moore (1986) [archived] examined 23 black/biracial children adopted by middle-class white families and 23 age-matched black/biracial children adopted by middle-class black families. When their IQ scores were measured (Table 2 [archived]), the black children adopted by black families had an average IQ of 104, while the black children adopted by white families had an average IQ of 117. The children were aged between 7 and 10 years at the time of testing.

Before exploring the important findings of this study, I must point out that the white families adopted a higher proportion of biracial children than the black families did. Therefore, one might argue that the 13 point IQ gap is caused by the differing degree of European ancestry of the two groups of children, rather than any environmental differences. However, after controlling for the race of the adoptive families, there is no IQ gap between black and biracial children. For children adopted into black families, the mean IQs of the biracial and black children were 105.7 and 102.9 points, respectively. For children adopted into white families, the mean IQs of the biracial and black children were 116.5 and 118.0 points, respectively. This suggests that the environment of the White families explains the higher IQ score for biracial children, rather than the other way around.

Now, I must be clear on what the important findings of this study are. It might seem like the important findings of this study are the high IQs of the black/biracial children. While these findings are interesting, these findings are not what important with respect to the race and cognitive ability debate. They are not important because, as the MTAS and other adoption studies have revealed, and as Rushton and Jensen (2005) have pointed out (page 259), IQ gains from adoptions tend to diminish over time. Therefore, we cannot assume that the IQ gains for these children will persist over time. Nevertheless, there are two sets of findings from this study that are important to the race and cognitive ability debate. The first is that the biracial did not have an IQ advantage over the black children. The second is that the children raised by white families had a mean IQ 13 points higher than the children raised by black families. These two sets of findings require an explanation. I will consider both in turn.

Firstly, after controlling for the race of the adoptive parents, there was no significant IQ advantage for biracial children. The IQs for biracial children was 2.8 points higher than “full” blacks within black families, and 1.5 points lower than “full” blacks within white families. How well are these findings explained by the genetic and environmental models?

Environmental model: the environmental model predicts these findings. The environmental model predicts that there will not be a large IQ gap between black children and black-white biracial children when raised in similar environments.
Genetic model: these findings contradict the genetic model unless one posits certain ad hoc assumptions (e.g., the black children were given superior environments compared to the biracial children, the black children were a genetically elite sample, etc.). The genetic model predicts that black children with higher proportions of European ancestry will have higher IQs than children with higher proportions of African ancestry when raised in similar environments.

In their section on racial admixture, Rushton and Jensen (2005) introduce this study as a study “seeming” to support the culture-only hypothesis (page 262). They then briefly describe the study and conclude with the statement “Given the young age of these children, a follow-up to adolescence would be informative” (page 262). I address the issue of young age of testing later in the post.

Secondly, black/biracial raised by white families had substantially higher IQs than black/biracial children raised by black families (gap of 13 IQ points). This held even after controlling for race. For biracial children, the mean IQ of the children raised by white families was 10.8 points higher than the mean IQ of the children raised by black families. For black children, the mean IQ of the children raised by white families was 13.1 points higher than the mean IQ of the children raised by black families. The most parsimonious explanation of these findings is that the environment associated with the white mothers is superior to the environment associated with the black mothers (e.g., differences in parenting practices, prenatal differences, etc.). How well are these findings explained by the genetic and environmental models?

Environmental model: the environmental model predicts these findings. The environmental differences that explain the black-white IQ gap in the general population are also present between the black and white families in this study.
Genetic model: these findings do not contradict the genetic model, but the genetic model has difficulty explaining these findings. As stated earlier, the most parsimonious explanation of these findings is that the environment associated with the white mothers is superior to the environment associated with the black mothers. The genetic model cannot explain why similar environmental differences are not responsible for a substantial portion of the black-white IQ gap within the general population.

Due to the two sets of findings described above, I conclude that this study provides strong evidential support for the environmental model.

In their section on transracial adoption studies, Rushton and Jensen (2005) quickly introduce and dismiss this study (and the Tizard study above) within one paragraph while noting “to be more informative, future studies need to be supplemented by follow-up testing, as in the Minnesota Study” (page 259). This comes after devoting several pages of discussion to the Minnesota Transracial Adoption study (pages 256-259). I address the issue of young age of testing later in the post.

Objections

Follow-up Testing

Many “hereditarians” have discounted the results from some of these studies because they did not perform follow-up testing on the children. For example, Rushton and Jensen 2005 state that one of the reasons that the Eyferth study is “ambiguous” is the young age of the children when tested (page 261).Further, they quickly introduce and dismiss the Moore and Tizard studies because of their lack of follow-up testing (page 259). Hereditarians argue that because the heritability of IQ increases as one ages, the IQ gains for black children may fade over time. While this is a valid criticism, there are several reasons why I believe the criticism is overstated:

The decline in IQ gains for the black children is not relevant. We are concerned with the IQ gap between blacks & whites. The Minnesota Transracial Adoption Study showed no significant change in the IQ gap after age 7, even though IQs diminished for all groups. The black-white IQ gap fluctuated by only about 2 points after age 7, and that fluctuation vanished after controlling for attrition. The authors even note “There were no adopted group differences in IQ decline” (page 124). Additionally, as far as I know, there is no independent data indicating that IQ diminishment is particularly strong for black adoptees compared to white adoptees when raised in enriched environments. Therefore, the lack of follow-up studies is no good reason to discount the results of these studies.
The heritability of IQ in young children is fairly significant. In reviews on intelligence research, Neisser et al. (1996) report that the heritability of IQ is 0.45 in childhood (page 85), Plomin et al. (2017) report a heritability of around 41% for children (age 9), and Plomin and Deary (2015) [archived] report that “The heritability of intelligence increases from about 20% in infancy to perhaps 80% in later adulthood”. Clearly, the heritability of IQ is large even in young children. Therefore, according to the genetic model, there should be a significant black-white IQ gap even among young childhood after equalizing environments.
Even ignoring the above points, the results from the Willerman and the Moore study are important not because they reveal anything directly important about genetics. They are important because they reveal something very important about the environments provided by white families/mothers vs black families/mothers. They reveal that the environments associated with white mothers/families are more conducive to cognitive development than the environments associated with black mothers/families. This finding is important regardless of the degree to which the children’s IQ is influenced by their genetics.

Another point is worth mentioning about this objection by Rushton and Jensen (2005). They examined three transracial adoption studies involving East Asian children – Clark & Hanisee (1982), Winick et al. (1975), and Frydman and Lynn (1989). These studies found that the East Asian children had mean IQs higher than the general white population (page 259-260). Rushton and Jensen take this to be evidence in favor of the hereditarian model since it suggests that the East Asian-white gap in cognitive ability is partially due to genetic differences (a hypothesis I will not address in this post). However, the children in these studies were also young at the time of testing. The children in Clark, Winick, and Frydman were aged 31 to 71 months (mean=44 months), 10 years, and 6-13 years (mean=10 years), respectively, at the time of testing. By comparison, the children in the Eyferth, Moore, and Tizard studies were aged 5-13 years (mean=10 years), 7-10 years (mean=8.6 years), and 2-5 years, respectively, at the time of testing.

Low sample size

Many hereditarians have criticized the studies favoring the environmental model because of their “low” sample sizes. For example, Kirkegaard et al. (2019) [archived] have referred to the Moore and Tizard as “tiny adoption studies”. Rushton and Jensen (2005) have also referred to these as “small sample studies” (page 259). While this is a valid critique, the critique is often not consistently applied. These hereditarians often use the Minnesota Transracial Adoption Study (MTAS) to support the hereditarian model, even though the sample size of the groups of interest (black and white children) in the MTAS is not particularly large. Consider the following sample sizes of the studies considered so far:

The MTAS has no advantage in sample size of white children. The MTAS had a sample size of n=25 (at age 7) and n=16 (age 17). By comparison, the Eyferth and Tizard (Table 5) studies had sample sizes of n=83 and n=39, respectively. The Moore and Willerman studies did not involve any white children.
The MTAS has no advantage in sample size of black children (children with two black parents). The MTAS had a sample size of n=29 (age 7) and n=21 (age 17). By comparison, the Tizard (page 343) and Moore (Table 2) studies had sample sizes of n=22 and n=26, respectively. The Eyferth and Willerman studies did not involve any black children.
The MTAS does have an advantage in sample size of black-white biracial children, but only compared to two studies. The MTAS had a sample size of n=68 (age 7) and n=55 (age 17). By comparison, the Eyferth, Tizard (page 343), Moore (Table 2), and Willerman studies had sample sizes of n=98, n=24, n=20, n=129 respectively. It’s not obvious that the sample sizes of the Tizard (n=24) and Moore (n=20) studies are “tiny” (at least not tiny enough to warrant ignoring their results) considering these are about the same size as the sample sizes of the black and white children in the MTAS (the two groups that hereditarian tend to focus on when analyzing the MTAS).

So the only real advantage that the MTAS has in sample size concerns the sample of black-white biracial children. But hereditarians must use the scores of the black and white children (which have low sample sizes) in the MTAS if they wish to cite the MTAS as evidence of the hereditarian hypothesis. There are a few reasons why they must do this:

If there is no comparison group for the biracial children, then their scores are pointless and cannot be used to provide support for the hereditarian (or environmentalist) hypothesis.
When interpreting the scores from MTAS, most hereditarians emphasize the gap in scores between the black and white children because those are the largest in magnitude.
Thomas (2017) has shown that the biracial-white IQ gap in the MTAS is only 2.5 ± 3.5 points, which is small enough to be “statistical noise”.

Therefore, anyone who appeals to the MTAS as evidence of the hereditarian hypothesis must use the scores of the black and white children, which have sample sizes that are no larger than the sample sizes of the other studies considered here. Thus, if one discounts the Moore and Tizard studies as evidence of the environmental model because of their low sample size, they must also discount the MTAS as evidence of the genetic model for the same reason.

Another point is worth mentioning about this objection by Rushton and Jensen (2005). They examined three transracial adoption studies involving East Asian children – Clark & Hanisee (1982), Winick et al. (1975), and Frydman and Lynn (1989). These studies found that the East Asian children had mean IQs higher than the general white population (page 259-260). Rushton and Jensen take this to be evidence in favor of the hereditarian model since it suggests that the East Asian-white gap in cognitive ability is partially due to genetic differences (a hypothesis I will not address in this post). However, two of these studies – the Clark (n=25) and Frydman (n=19) studies – had sample sizes that were lower than the sample sizes of the Moore (n=46) and Tizard (n=85) studies. It is strange that they labelled the Moore and Tizard studies as “small” (page 259) without applying a similar label to the Clark and Frydman studies.

Conclusions

As far as I know, that’s all the data available on the cognitive ability of black children raised by white caretakers. Even if you disagree with my interpretation, this post should be valuable as a coherent collection of all the relevant studies along with their biggest criticisms. Of course, many of these studies have significant problems (e.g. low sample sizes, lack of follow-up testing, etc.) which means that no interpretation is conclusive. Further data is needed to justify more confident conclusions. That being said, I believe that the most reasonable assessment of all the available data suggests the following main takeaways.

1. The preponderance of this data provides most evidential support for the environmental model

Consider the following main findings from the previously cited studies that require explanation:

The MTAS found that white adoptees outscored biracial adoptees who outscored black adoptees. The environmental model can explain these patterns if one makes certain reasonable but unverified assumptions (e.g., that the black children had inferior pre-adoptive environments) whereas the genetic model can explain the racial gaps without making any further assumptions. Therefore, the genetic model better explains these findings.
The MTAS found that the IQ gap between blacks and whites did not grow between ages 7 and age 17. The environmental model is consistent with these findings, but it neither predicts nor explains these findings. The genetic model is contradicted by these findings unless one posits certain ad hoc assumptions (e.g., the environment for the black children improved over time to offset the increasing influence of their disadvantaged genes). The genetic model predicts that the IQ gap would widen as children age (assuming they are raised in similar environments) because the heritability of IQ increases dramatically as children age. Therefore, the environmental model better explains these findings.
The Eyferth study found no IQ advantage for children with higher levels of European ancestry than African ancestry. The environmental model predicts these findings. The genetic model is contradicted by these findings, unless one posits certain ad hoc assumptions that cannot be verified (e.g., the black children were a genetically elite sample, the black children had genetically elite mothers, etc.). Therefore, the environmental model better explains these findings.
The Kirkegaard study found no IQ advantage for children with higher levels of European ancestry than African ancestry. The environmental model predicts these findings. The genetic model is contradicted by these findings, unless one posits certain ad hoc assumptions that cannot be verified (for similar reasons as explained in finding 3). Therefore, the environmental model better explains these findings.
The Tizard study found no IQ advantage for children with higher levels of European ancestry than African ancestry. The environmental model predicts these findings. The genetic model is contradicted by these findings (see finding 3). Therefore, the environmental model better explains these findings.
The Willerman study found that biracial children raised by white mothers had a mean IQ 9 points higher than biracial children raised by black mothers, even when the two sets of parents/mothers had similar SES backgrounds. The most parsimonious explanation of these findings is that the IQ gap between these two groups of black children is due to differences in environments produced by white mothers vs black mothers. The environmental model predicts these findings; the environmental differences that explain the black-white IQ gap in the general population are also present between the black and white mothers in this study. The genetic model, while not directly contradicted by these findings, is undermined by these findings, unless one can explain why the environmental differences that produce the IQ gaps in these studies do not also produce the IQ gap in the general population. Therefore, the environmental model better explains these findings.
The Moore study found no IQ advantage for children with higher levels of European ancestry than African ancestry. The environmental model predicts these findings. The genetic model is contradicted by these findings (see finding 3). Therefore, the environmental model better explains these findings.
The Moore study found that black/biracial children raised by white families had a mean IQ 13 points higher than black/biracial children raised by black families. The environmental model predicts these findings. The genetic model is undermined by these findings (for similar reasons as explained in finding 6). Therefore, the environmental model better explains these findings.

The problems mentioned in these points for the genetic model also apply to the intermediate model. Thus, the preponderance of data seems to provide most evidential support for the environmental model. For 8 of the 9 findings, the environmental model outperforms the genetic model. There is only 1 finding that is better explained by the genetic model.

Note, however, that while these findings are most easily explained by the environmental model, this data does not rule out the genetic model or intermediate model. This is because, as stated earlier, much of the data here is flawed for a variety of issues (e.g., possible sampling error, low sample sizes, lack of follow-up testing, etc.). An ideal study would involve collecting large representative samples of black and white children, raised in a representative sample of adoptive white families, where the IQs of the biological parents of the black and white children are known, with regular follow-up testing of the IQs of the children.

2. The environmental model explains cognitive disparities at childhood and achievement disparities at adolescence

One might remain skeptical that the environmental model explains the racial IQ gap at adolescence. Perhaps one discounts the studies from Eyferth, Kirkegaard, Tizard, Moore, and Willerman because of a lack of follow-up testing (despite my responses to this objection above). Even accepting this, the data still indicates that the environmental model explains the following gaps:

The IQ gap at childhood (according to all studies other than the MTAS),
The academic achievement gap at adolescence (based on results from the MTAS). This is especially true for vocabulary and reading achievement, but is also true for mathematical achievement to a large degree.

3. There is little practical importance in continued arguments about the genetic component of the black-white IQ gap

Until we reach black-white parity in adolescent academic achievement and in childhood IQ/achievement, we know that there are still significant environmental gaps regarding intellectual development in these areas. It is only practically important to quantify the genetic component of the IQ gap if we need to know when we have reduced all significant environmental disparities. But we know we haven’t reached that point yet because there are still substantial black-white disparities in childhood IQ and adolescent vocabulary/reading achievement (as I indicated in a previous post, the average black 17-year-old is less proficient in reading than the average white 13-year-old) which are driven largely by environmental factors (based on the studies here). We can postpone investigating the precise genetic component of the IQ gap until we have eliminated these known significant environmental differences.

To be clear, I only mean to say that there is no practical importance to precisely determine the genetic component of the black-white IQ gap. There may still be academic or scholarly importance to perform this endeavor.

4. It is unclear what factors constitute the important environmental differences between blacks and whites

These IQ studies show that there are significant environmental differences between Black children raised by Black parents versus White parents. The differences persist when we control for maternal marital status and when we limit to middle-class families. In my post on candidate explanations of the cognitive ability gap, one promising candidate suggested by the studies I considered was differences in parenting practices. In fact, Moore (1985) noted that middle-class Black adoptive mothers were much more reprimanding and scornful while observing their children solve problems, whereas middle-class White mothers were much more encouraging and rewarding during the same task. Of course, more data is needed to make any conclusions.

It should also be noted that even if differences in parenting practices explain most of the IQ gap, it is still possible that differences in parenting practices are themselves the result of genetic differences. Most human behaviors are highly heritable, so parenting practices are probably heritable as well. Thus, it’s possible that genetic differences in, say, maternal sensitivity explain the IQ gap. While this is certainly possible, it is not possible to verify or falsify this claim given the current data (I’m also not sure how this thesis could be feasibly tested).

Related Works

Ulric Neisser et al. (1996). “Intelligence: Knowns and unknowns” [archive]. Inspired by the heated debate regarding intelligence following the release of The Bell Curve, the Board of Scientific Affairs of the American Psychological Association established a task force of 11 experts on intelligence to prepare an authoritative report surveying the current state of the field. The report was continually revised and discussed until the report received unanimous support from each member of the task force.
Christopher Jencks and Meredith Phillips (1998). The Black-White Test Score Gap. A comprehensive collection of works outlining the history, causes, and impacts of Black-White test score gaps. The first chapter can be found here [archive].
John C. Loehlin (2000). “Group differences in intelligence“. Handbook of intelligence. Loehlin much of the same evidence cited here and comes to the conclusion that more research is needed to come to a conclusion on whether genetic differences are responsible for the cognitive ability gap.
J. Philippe Rushton and Arthur R. Jensen (2005). “Thirty years of research on race differences in cognitive ability” [archive]. Rushton and Jensen review much of the data cited in my post as evidence for the “hereditarian” model of the cause of the black-white gap in cognitive ability.
Richard E. Nisbett et al. (2005). “A Commentary on Rushton and Jensen” [archive]. An environmentalist objection to the argument from Rushton and Jensen’s article “Thirty years of research”.
J. Philippe Rushton and Arthur R. Jensen (2005). “Wanted: More Race Realism, Less Moralistic Fallacy” [archive]. A response to the objections from Nisbett in “A Commentary on Rushton and Jensen” and other environmentalist criticisms.
Nicholas Mackintosh (2011). “Group differences”. IQ and Human Intelligence (page 148-182). Mackintosh reviews much of the same evidence cited here, and comes to the conclusion that the data “provide little support for the genetic hypothesis” (page 152).
Richard E. Nisbett et al. (2012). “Intelligence: New Findings and Theoretical Developments” [archive]. An authoritative review of the field of intelligence to update the “Intelligence: Knowns and unknowns” (1996) article. This review was authored by 6 experts who surveyed many of the facts discovered since the publication of the original article.
Drew Thomas (2017). “Racial IQ Differences among Transracial Adoptees: Fact or Artifact?” [archive]. A critical analysis of currently available transracial adoption studies, addressing some methodological flaws of previous interpretations of the data.

For arguments from prominent advocates of the environmentalist model, see works by Richard Nisbett and James Flynn. For arguments from prominent advocates of the hereditarian model, see works by Philippe Rushton and Arthur Jensen.

Reason without restraint