The following points were critically raised by Turkheimer and Harden (2024) and the relevant page number from their article given. My responses follow directly after each criticism.

(1) Using a map from Biasutti (1967, p. 224, “distribuzione della varia intensita del colore della pelle” [distribution of various intensities of skin color]) taking skin lightness as an indicator for ancestry; transfer of hatching on a map into numbers for countries by a student (p. 2).

The answer is already in the manuscript, and can be found in the article’s supplementary materials (Rindermann et al., 2024, Appendix, p. 5):

There are two points of criticism:

Of course, controlling for societal-cultural factors is advisable in statistical analyses. For an alternative measure of ancestry, see the next point.

(2) Using a map from Beals et al. (1984, p. 304, “Brain size, cranial morphology ...”) as source for world-wide distribution of brain size (“cranial capacity”); transfer of hatching on a map into numbers for countries by a student (p. 2).

There are two points of criticism:

But there are also other useful indicators of ancestry developed by colleagues: Genetic distance from the cradle of Homo sapiens in South Africa or Ethiopia. Cavalli-Sforza et al. (1994, pp. 75–76) calculated F_ST genetic distances between 42 populations. Spolaore and Wacziarg (2009, pp. 480–487) used these data to calculate the genetic distances between countries at the country level and their current populations. F_ST distance is first and foremost a measure of the genealogical relationship between populations; if one takes South Africa as the “cradle of humanity” as a starting point (León & Burga-León, 2015), then it represents an indicator of ancestry as evolutionary history.4

Upon closer examination, the data revealed some implausible results: the four countries that were genetically most similar to South Africa were Algeria, Libya, Morocco and Tunisia; only then came Niger and other countries south of the Sahara. The population of the four North African countries’ neighbor, Egypt, is genetically much more distant from the South African population (which is plausible) than that of these four countries. I wrote to Romain Wacziarg, who pointed to a more recent and better data set on genetic distances (Spolaore & Wacziarg, 2018). These data are based on human microsatellite variation from Pemberton et al. (2013) covering 267 worldwide populations. Again, Spolaore and Wacziarg calculated the genetic distances between countries at the country level and their current populations.5 I used these data to take the genetic distance from Ethiopia (also the cradle of humanity).6

In an international sample, the correlations between genetic distance to South Africa and Ethiopia and the g-factor ancestry used in the article (Rindermann et al., 2024) were r = .69 and .40 (N = 174 and 159 countries, respectively; the two genetic distance measures correlated r = .67, N = 172). In the 15-country sample with at least five BOMAT IQ (a culture reduced figural test of reasoning) observations per country, the correlations with the g-factor ancestry were r = .72 and .93 (N = 15 countries). If we replace the g-factor ancestry with the genetic distance measures for Table 5 of the article, the correlation with individual BOMAT-IQ changes from r = .29 (N = 497) to r = .22/.28 (N = 499/497). When controlling for geographical distances between countries, the effects of genetic distances remain (β = .16/.44, South Africa/Ethiopia) and are considerably larger than the geographical ones (β = .08/−.22; N = 499/497). For Table S1 (Correlations at the country level; Rindermann et al., 2024) the correlation with country BOMAT-IQ changes from r = .67 (N = 15) to r = .57/.77 (N = 15). Similarly, when controlling for geographical distances between countries, the effects of genetic distances remain (β = .32/.84, South Africa/Ethiopia), but they are not always larger than the geographical ones (β = .40/−.08; N = 15).

Compared to the g-factor ancestry the correlations are slightly smaller or higher. In principle, the relationship between ancestry factors is confirmed. To use this variable in a path model, see below. Of course, genetic effects must always be controlled for possible environmental factors (e.g., see above, geographical distances). But even when these are controlled, genetic effects (on intelligence) remain (see the studies from Becker and Rindermann (2016), Kodila-Tedika and Asongu (2016) or Rindermann et al. (2012)).

Again: All the data used here are suboptimal. The exact sources for skin lightness and cranial capacity are unclear, and there is obviously some implausible data on genetic distances. Data for individuals would be better still. However, all this is likely to lead to an underestimation of possible genetic effects. Finally, the use of genes that encode intelligence and causally influence it (via neurological processes and structures) would be optimal. Everything else provides only indirect, tentative evidence.

Another relevant genetic effect (not mentioned in our study) is exogamy vs. marriages between relatives (consanguinity). Consanguineous marriages have harmful health effects, including neurological and psychological implications. (The fate of the Spanish branch of the Habsburgs is notorious.) Exogamy also entails—from today’s perspective—positively valued social, political and cultural developments such as rights to individual self-development and democracy (Henrich, 2020; Schulz, 2022; Woodley & Bell, 2013). Presumably this tradition or its rejection goes back to cultural factors, Christianity plays an important role (e.g., Henrich, 2020; Rindermann, 2018, pp. 313–316), but exogamy is even older in Europe (e.g., MacDonald, 2019). In contrast, consanguineous marriages are common in Arab, Muslim and African countries. Using country level data from Woodley (2009) and Tadmouri et al. (2009) the correlation of consanguinity with individual Bomat-IQs is r = −.09 (N = 331) and at the country level r = −.17 (N = 9). The correlations are not high but negative as assumed. Data are missing for several countries, and individual data for each migrant would certainly be better.

(3) Using a factor analysis with two variables (“One of the most basic rules of factor analysis is that any factor solution requires at least three indicators.)” (p. 2).7

Turkheimer and Harden (2024; Turkheimer, 2024) assert firmly that factor analysis requires at least three variables. They do not provide any reasons or references for their assertion and, furthermore, their claim is not correct: In a factor analysis with two variables the loadings of the variables must both be fixed to 1 (assuming essential tau equivalence). The model is then saturated (df = 0; “just identified”; e.g., Geiser, 2023). The “holy book” of statistical methodology by Jacob Cohen and colleagues (Cohen et al., 2003, p. 473, Figure 12.5.2) gives an example of this (a latent variable with exactly two indicators), as does the more recent Backhaus et al. (2023, p. 417, their Figure 7.4). However, we have used here a principal component analysis for extraction. This also works without such restrictions with two variables. Nevertheless, the scientific gain compared to a simple item averaging (after z-standardization) for a scale is zero—the g-factor and the mean of such an item averaging correlate at r = 1.8 This also proves that the g-factor construction via the first unrotated factor is correct, but unnecessarily complicated.9

(4) In parentheses of the path analyses shown, correlations are given rather than standard errors (p. 3).

We explained this in the published paper (Rindermann et al., 2024, p. 16):

Standard errors were reported in the publication in Tables 2 and S2 (Rindermann et al., 2024).

(5) Standard errors are not interpreted, no use of significance tests (p. 3).

Standard errors as well as significance tests are both part of the idea of inferential statistics, which is generally not very useful, especially at country level (Pollet, 2013). As written in the original manuscript, here extended by two more references and the publication titles:

Significance tests were not used for interpretation (for a detailed rationale, see, e.g., Cohen (1994), “The earth is round (p < .05)”; Falk and Greenbaum (1995), “Significance tests die hard. The amazing persistence of a probabilistic misconception”; Gigerenzer (2004), “Mindless statistics”; Hunter (1997), “Needed: A ban on the significance test”; Wasserstein et al. (2019), “Moving to a world beyond ‘p < .05’”). More instructive for inductive generalization—which is not possible with significance tests—is the demonstration of the stability of relationships and effects across different country samples, different variables, different measurement points and various studies by different authors using different methods. (An attempt in this direction was already the analysis with three different data levels in the study—individual, multilevel, country average; Rindermann et al., 2024, Figures 2, 3, S2.)

The larger a standard error (or “standard uncertainty”), the less one can trust the exact value of a path coefficient (beta), the larger the range around the given value in which the true value can lie (beta coefficient minus and plus standard error). If this range includes 0, the true effect could also be zero (meaning no effect).

(6) Warning messages in the Mplus output indicate that standard errors may not be trustworthy (p. 3).

There were warning messages for the path model depicted in Figure 3 (Multilevel path analysis at the individual and country level …; Rindermann et al., 2024): “The model estimation has reached a saddle point or a point where the observed and the expected information matrices do not match. ... The standard errors of the model parameter estimates may not be trustworthy for some parameters ... The nonidentification is most likely due to having more parameters than the number of clusters. Reduce the number of parameters.” However, when considering the standard errors (Table S3 of the article’s appendix; Rindermann et al., 2024), no conspicuous values were found. Nevertheless, the effect of GDP was unusually high (β_GDP = .70). The implications of these warnings are difficult to assess. This requires comparative analyses with different variables and numbers of variables, which we do below. In addition, there are correlational results and traditional regression analyses that confirm the findings.

As quoted, Mplus recommended a reduction of the parameters. We removed the variable with an effect outside the usual range of values (GDP). Now the Mplus warnings disappeared. The effects remained the same except for the increase in the direct impact of the country cognitive ability level (from β_CA = .36 to β_CA = .79; see Figure 1 here). Although the model is now without warnings, it could be “misspecified”, i.e., a potentially important variable or determinant (here: wealth) is missing. To have both (models without warnings and with GDP) we would need larger country samples (more than 15 countries as given here).

But there has also been criticism of the ancestry factor (see above). If we replace this with genetic distance from the cradle of humanity (Ethiopia; see above), the effects change only minimally. Behind cognitive ability were ancestry (β_Anc = .69, now β_DenD = .67) and culture (β_Cul = .29, now β_Cul = .38; see Figure 2 here).10

We have now tried to take into account the individual income in the country of origin instead of the level of societal wealth (logarithm of average GDP per capita). However, the warning messages appeared again, the same as for Figure 3 of the article (Rindermann et al., 2024; from the Mplus output: “Observed and the expected information matrices do not match. ... Reduce the number of parameters.”). In addition, there was now a standard error outside the defined range of values (see Table 1; SE = 1.34 for the effect income → BOMAT-IQ). If it is above 1 for standardized effects, it is undefined (only suppressor effects may make sense). The results should therefore be interpreted with caution.

Plausibly, individual income depends on one’s own education (β_School = .52; see Figure 3 here) and country cognitive ability level (β_CA = .73; GDP/c would also have to be included here). The effect of income on individual intelligence is considerably smaller (β_Inc = −.03) than that of education (β_School = .36), which is consistent with similar findings regarding parental education and parental income effects on the IQ of their children (Rindermann & Ceci, 2018). Due to the warnings, in a further multilevel model we only used the three individual variables, but the pattern of results is repeated: Education (β_School = .39) is more important for intelligence than income (β_Inc = .08). Education has a positive impact on income (β_School = .34).11 However, due to the small number of countries (and the warnings for Figure 3), the findings should be interpreted with caution and require replication in larger samples.

And the following limitation applies to all models here: Reciprocal effects between intelligence and education as well as intelligence and income were not taken into account. This means that the models provide only a partial representation of reality.

(7) Poverty, political strife, war (p. 3).

Turkheimer and Harden (2024, p. 3) suggested that poverty, political conflict and war, in short, experienced stressors, are responsible for the weak test results (relative to the First World). “Some refugee immigrants, fleeing from countries beset by poverty and political strife, now torn apart by unrelenting war, obtain upon arrival in Germany lower scores on IQ tests than the resident German population.”12 This is a plausible hypothesis. According to DeGroot (1951), intelligence in the Netherlands fell during the Second World War and the Nazi occupation from an IQ of 101.5 to 97.7, i.e., by around −4 IQ. We have already dealt with this aspect in the study. Based on the data, it is probably not the case that such experiences are decisive for this result. As we discussed in the article, the IQ results of refugees are better than the those found in their country of origin (usually collected before armed conflicts):

Another study on people who had experienced most extreme stress (Holocaust survivors; meta-analysis by Barel et al., 2010) pointed to stress (posttraumatic stress symptoms, d = .72; their Table 6), but cognitive effects were low (cognitive functioning including intelligence, d = .10; their Table 8). Yes, the assumption of stress-reduced intelligence is plausible, but if empirical studies do not confirm this, then this thesis can no longer be defended. (For the effect of political conditions, see point 9 and Section 4.2.1. here)

Finally, it is not just that “some refugee immigrants” score poorly on cognitive tests, but this is the average of four studies (see another below).

(8) Study results are not sufficiently controversial. No legitimate controversy (p. 4).

Turkheimer and Harden (2024, p. 4) question on the one hand whether the refugee IQ study is sufficiently controversial (“Is this controversial?”), but on the other hand they do not rate the study as a “legitimate controversy”. So the entire paper is somehow wrong for the Journal of Controversial Ideas. Strange. It seems that, from their point of view, no such study should be published at all.

(9) Cognitive ability treated as an “essential feature baked into immigrants, an ultimate cause of the chaos in their home countries and fraught migration to new homes in Europe. The skills tested by a matrix reasoning task, we are told, are the ultimate cause of wealth and democracy, rather than their result. There is, however, no evidence this is the case” (p. 4).

What do Turkheimer and Harden mean by an “essential feature”? If they mean an unchangeable feature, then this is not our position. Our own studies (and those of other authors) have also shown the changeability of intelligence and student achievement in countries at different stages of development (for Turkey: Rindermann et al., 2013; USA: Rindermann & Thompson, 2013; worldwide: Rindermann & Becker, 2023). We also do not assume one-sided effects of intelligence. Many of my studies in particular have analyzed interactions between intelligence and society. For example, between cognitive human capital (IQ), economic freedom and wealth (income): Wealth is more relevant for the cognitive development of poorer countries and cognitive ability is more relevant for the wealth development of richer countries (Rindermann, 2018, pp. 225–233). The same holds for political conditions (“positive politics” which includes the rule of law, political freedom and democracy): While cognitive ability showed in cross-lagged panel studies a strong positive impact on politics, the effect of politics on cognitive ability was zero or close to zero (Rindermann, 2018, pp. 244–248).

Overall, however, many research results and, in particular, statements by scientists in this field remain contradictory. For example, there are well-known scientists who claim that education has no effect on cognitive competence. Education is claimed to be only an indicator of ability (Spence, 1973; Charlton, 2009; G. Miller, 2009; Murray, 2008; Caplan, 2018). As an actor and recipient of research, one should in the end take a step back, intellectually allow a variety of methods, consider the work of different authors from different disciplines and paradigms and come to a balanced judgment.

The style and tone of Turkheimer and Harden’s commentary are unusual for a scientific article. Quotes from the text:

And in a commentary published by Turkheimer (2024) alone on Substack:

The style of these comments could signal that something is wrong here. Scientists engage in argumentative discussions with others. In such discussions, excessive emotion is not helpful. Credible and reasonable thinkers qualify their claims, consider and appreciate reasons that point in a different direction, and tend to take the viewpoints of others seriously.

Turkheimer is a co-author of a recently published article (Matthews et al., 2024) that accused other scientists of conducting “abhorrent science” and suggested excluding such studies from the publication process. In an earlier article, Turkheimer appealed to a supposed ethical principle (which reflected political–ideological criteria) to determine which empirical claims should be accepted:

Bernard Davis (1978) coined the term “the moralistic fallacy” for answering scientific questions through moral-political values. Ultimately, the pretense here is to think morally, but this actually serves ideological power politics: fighting those who think differently and excluding their views from publication. The contributions by Turkheimer and colleagues make it clear that we need an ethics of behavior among scientists, an ethics of argumentative debate and an ethics of dealing with others in general (e.g., Rindermann, 2006). Empirically, contributions could be evaluated by independent raters with respect to their potential for hate or incitement, and these evaluations could be examined in relation to the empirical and innovative content of the contributions.

Two philosophers have critically noted that the intelligence of refugees is not a relevant criterion for determining whether there is an ethical obligation to help them. Even so, Christopher Heath Wellman (2024) discusses whether relatively low intelligence does not lead to greater burdens on the part of the host country as well as the immigrants and whether it is therefore advisable to help people in their home countries and regions rather than supporting them in their resettlement. Another aspect is that when the relatively more intelligent people of a country emigrate, development in the country of origin is impaired. African countries, but also Venezuela, are suffering from brain drain (Kapur & McHale, 2005).13 Some have proposed a kind of tax to be paid by the emigrating skilled workers who benefit from brain drain (“Bhagwati tax”; e.g., Oldman & Pomp, 1975). But the idea is already around 50 years old and still not realized and probably never will be, not even in the form that countries that benefit from the brain gain would pay for it (e.g., Switzerland; Ben-Hur et al., 2017). However, the advantages of emigration of cognitive elites (“intellectual classes”; Coyle et al., 2016) are also pointed out: if they work in wealthy countries, they send money to those who stayed at home, which generates more income and wealth than development aid. In addition, this motivates people in the relatively poorer emigration countries to invest in education. For example, in Pakistan, Naipaul (1981) described a mother and member of a religious minority (followers of the Ahmadiyya doctrine, which is related to Islam) who was particularly concerned that her children learn English so that they could go abroad. The possibility of emigration also influences politics in the home country, reducing the likelihood of competent citizens being too badly treated.14

Bradley Hillier-Smith (2024) argues that the obligation to support refugees does not depend on their cognitive characteristics (pp. 6–8) and the costs involved (pp. 3–6, 14–16).15

Bradley Hillier-Smith (2024) used the term “obligation” 73 times in his commentary; five of them appear in the abstract, and once in the actual title: “obligations to refugees”. In addition, he speaks of “duty” and “obligations” six times, e.g., as “positive duties owed to refugees”. But not once does he mention the duties of migrants or refugees toward the host society or the duties of politicians toward their citizens. The book he refers to by D. Miller (2016), “Strangers in our midst”, is similar: There is a chapter on “The rights of immigrants”, but no chapter on the rights of the host society and the people living there. But, as Hillier-Smith mentions, the rights of all humans count: “It has been long- and well-established that all persons have equal moral worth” (Hillier-Smith, 2024, p. 6).

In addition, Hillier-Smith (2024) emphasizes several times that the cost of protecting the refugees is minimal for the host population: “at relatively little cost” (three times on p. 12). And: “There is a widely accepted robust moral constraint against harming or violating the rights of innocent people (indeed, it is perhaps the most fundamental commitment in common morality)” (Hillier-Smith, 2024, p. 12). He can rely on an influential article by Peter Singer:

In his section 4 (Hillier-Smith, 2024, pp. 9–13) he describes in detail and emphatically the often harsh and terrible fates of refugees, their situation in their countries of origin, during their flight and in the host country. In contrast, he does not mention the fate of the people in the host countries, nor does he address their rights. And he writes again and again about “little costs”. For some, however, these “little costs” mean the loss of their lives. There are many, many sacrifices here of comparable moral significance. If Hillier-Smith describes the fate of migrants over many pages, which I and we all should acknowledge, then it should also be permissible to describe the sacrifices of the people in the host societies, and it is morally imperative that these sacrifices are acknowledged and recognized:25

There have been many, many similar incidents, which I will include, for those interested in the shocking details, in an Appendix A. All of this and every single event is shocking, intolerable, completely unacceptable and in no way justifiable as a situation brought about in Europe. Victims are not only the dead and injured, but also their relatives, paramedics, police officers, observers and all those and their relatives who could have become victims and only did not become victims by chance. These are not “little costs” and they are also “violating the rights of innocent people (indeed, it is perhaps the most fundamental commitment in common morality)” (Hillier-Smith, 2024, p. 12).

These developments were not unforeseeable. Something similar was described decades ago in dystopian articles and books, but in milder versions that underplayed the coming horrors. Their authors became controversial and therefore—so it was thought—could be ignored.27 A more journalistic piece also came from me (Rindermann, 2015).28 What has happened is even worse than predicted: the idea of someone driving a vehicle through crowds at festivals used to be unthinkable.

Of course it is possible to put forward counter-arguments. Perhaps one or the other number is biased. As mentioned, a higher proportion of refugees are young and male. Of all people, 99%, including young male refugees, are peaceful and friendly for 99% of their lifetime. For 99% of the victims’ lives, they were not victims. It is often pointed out that the acts were committed by people who had mental health problems (including some of the examples given in the appendix). Although the risks of violence are indeed increased in certain mental illnesses (this applies in particular to schizophrenia spectrum disorders), mental illnesses can explain only a very small fraction of general violence (Ahonen et al., 2019; Whiting et al., 2021). Certainly, there has been violence in football in Europe before, for example in 1998 when a French policeman was kicked and punched in the head by German hooligans. He was left with a permanent disability. “The attack was deemed a national shame by Chancellor of Germany Helmut Kohl. In the years since, [the French policeman] Nivel has been a guest of honor at Germany football matches and a foundation in his name has opposed football hooliganism.”29 Have we seen such a reaction in the case of immigrant violence?

But all these objections are irrelevant. Existing violence is no justification for letting immigrants increase it further. Biased (racist etc.) reporting does not mean that there is only a negligible amount of crime. These objections are misused as strategies for refusing to perceive reality.

The costs of immigration are not only economic. Other concerns of the host countries and their populations, the legitimate interests of their people, must be recognized. Does Hillier-Smith acknowledge them, indeed does he perceive them at all? Just as we recognize and empathize with the fate of some who seek to flee their countries of origin, so we must also recognize the fate of some residents of the receiving countries.

At the center of the moral controversy over whether a country should accept refugees is that the rights and interests of different people are in competition with each other. The legitimate interests of the present residents of the receiving country are in competition with the legitimate interests of the migrants.

The usual reactions to this problem are to ignore it (to pass over it with silence), trivialize it (it’s not all that bad), twist it into the positive (“enrichment”) or stigmatize those who point it out (evil, stupid, racist). For example, the leading German politician of The Left party,30 Gregor Gysi, commented on the numerous murders recently committed by migrants in Magdeburg, Aschaffenburg and Munich: “Murders by Germans also happen here all the time.”31

The classic ethical answer is: All humans are equal. “It has been long- and well-established that all persons have equal moral worth.” (Hillier-Smith, 2024, p. 6) Thinking and acting here is guided by universal ethical principles that encompass all of humanity. An example could be the behavior of the parents who, after the murder of their daughter (who is said to have been also active in refugee aid) by an Afghan refugee, established a foundation that supports students, including foreign students and students with refugee backgrounds, with a focus on development aid and student internships in developing countries.32 These parents are taking care of others, investing resources, money and time, so that people who are not part of their family are supported. This is exactly according to the model of universalist ethics that treats all people equally. However, this is fundamentally at odds with widely practiced rules of morality. In a sample of 60 different societies, the following ethical orientations and practices were found to be widespread worldwide (Curry et al., 2019):

At the heart of morality here is cooperation, for the benefit of all involved, which has developed because morality as cooperation benefits those who practice it and their offspring. In contrast, it can be argued that to practice purely universal morality will be harmful in the long term; or even that it is immoral because it places excessive burdens and risks on those close to us and those who depend on us.

The refugee policy currently practiced in Germany and some other countries is contrary to these widely practiced rules of morality. The host communities concerned are inflicting costs on themselves, and the policy has terrible, harrowing and fatal consequences for some of their members. In our political and private actions, we must be guided by sustainable and long-term viable principles. This also applies to politicians, who in many countries are constitutionally obliged to ensure the welfare of the citizens and residents of their country.34

This does not mean that the claims of others, even those distant from us, are null and void; however, these claims must be balanced against one another and more strongly oriented toward the well-being of those for whose existence we are responsible. We need an ethic of personal responsibility—an ethic of commitment toward those who would not exist without us, whose well-being we are responsible for, and who depend on us—then toward our partners, parents, relatives, friends, neighbors, and loved ones, who would also help us in times of need (reciprocity). Expanding the circle of those whose interests are to be considered does not release us from our responsibility toward those close to us and dependent on us; and in considering the interests of others, we must in no circumstances endanger their life and well-being.