The 2023 Cremieux Reared-Apart Twin IQ Figure: Pretty Colors, Nothing There

A recent critical review of Thomas J. Bouchard’s “Minnesota Study of Twins Reared Apart” (MISTRA) by David Bessis led me to take a closer look at a 2023 figure by @cremieuxrecueil (AKA Jordan Lasker). The figure supposedly shows that the “average absolute difference in IQ” between allegedly “reared-apart” MZ (identical) twin pairs (known as MZAs) is similar to the IQ score difference of the same person taking the test twice, and that these differences are much smaller than the 17-point difference between “two random people.” Bouchard et al. made this claim in their 1990 Science Magazine article. The implication is that MZA data shows that heredity strongly influences how people score on IQ tests.

That’s basically what Cremieux was trying to reproduce in this difficult-to-decipher figure. The multi-colored vertical bars are just a scientific-looking distraction from the above claim. In this brief post, I will show that the claim is seriously flawed and that the figure is based on studies of twins supposedly “reared apart,” but who were not reared apart using any reasonable definition of the term. In addition, the figure omits key control-group data.

I present the Cremieux figure as it was reproduced in 2023 by Paul Graham, who has over two million followers on “X” (Twitter). As Bessis noted, Graham’s post had been viewed over 3.5 million times.

The Cremieux figure was based on a figure published in the MISTRA 1990 Science article, and a 1970 table by IQ hereditarian psychologist Arthur Jensen. As I will explain, there are several reasons why the Cremieux figure should be disregarded.

1. The (yellow) Burt results are based on a discredited study. Cremieux attempted to rehabilitate the long-discredited “twins reared apart” (TRA) study by the British psychologist Cyril Burt. (Cremieux did provide another figure without the Burt data, where he arrived at similar conclusions.) The story of the “Burt affair” has been told hundreds of times since the 1970s, including in Stephen J. Gould’s The Mismeasure of Man. Most authors who defended Burt against charges of fraud acknowledged that his results lacked scientific merit and cannot be trusted. Articles published in 1994 and 1997 by the late Princeton psychologist William H. Tucker provide a definitive explanation for why Burt’s “data” must remain stricken from the scientific record.

2. The 1937 Newman et al., 1962 Shields, and 1965/1980 Juel-Nielsen studies (respectively, blue, purple, green) were based on MZA twins only partially reared apart. For evidence, see my 2020 Mad in America post based on information I compiled for Table 2.1-.2.3 of my 2015 book, The Trouble with Twin Studies. The best way to approach Tables 1-3 in my 2020 post is to read the case history information in the right-hand “From the Case Descriptions” column from top to bottom. As readers will see, most of the 75 pairs described in the Newman et al., Shields, and Juel-Nielsen studies had contact and shared many environmental features. The detailed case history information they left behind shows that, in addition to sharing a prenatal environment:

· In 25/75 (33%) of the pairs, twins were separated at 12 months of age or later.

· In 56/75 (75%) of the pairs, twins had contact with each other while growing up.

· In 42/75 (56%) of the pairs, one or both twins were placed with a family member.

· In 17/75 (23%) of the pairs, twins lived together for at least 12 months after separation or grew up close by to each other.

People calling these 75 pairs “reared-apart identical twins” either (1) have not read the original studies’ case descriptions, (2) have a radically different idea of what it means to be “reared apart” than most other people have, or (3) are trying to deceive uninformed readers and audiences.

3. The MISTRA (orange) results are also based on twins only partially reared apart. The Minnesota researchers have consistently refused to grant independent analysts access to their unpublished raw data and case history information. Therefore, we can safely assume that the similarity-biased MISTRA MZA pairs were no more “reared apart” than were the MZA pairs in the three earlier studies seen in the Cremieux figure. Conclusion: No study of truly reared-apart twins has ever been performed. What we have, as psychologist Susan Farber concluded long ago, are studies based on partially reared-apart MZ twin pairs.

4. The “two random people” IQ difference mean score is misleading. MZAs are the same age and sex, usually grow up in similar socioeconomic (SES) conditions, and share many other environmental factors and experience similar age-cohort influences. Therefore, unless pairs of “random people” are matched on age, gender, SES, and many other environmental factors, their mean IQ score difference cannot be compared with that of MZAs.

The IQ score similarities of MZA twins are also influenced by the “Flynn effect.” Moral philosopher and IQ researcher James Flynn showed in the 1980s and 1990s that IQ scores worldwide had been increasing by about three points per decade (0.30 points per year), including supposedly “culture-fair” tests such as Raven’s Progressive Matrices (increasing 0.50 points per year worldwide). Therefore, the educational, learning, and cognitive skills development of MZA pairs began at the same Flynn effect starting-point in similar era- and location-related “huge IQ advantage or disadvantage” environments. Contrary to the key TRA study assumption that an above-zero MZA IQ correlation is caused solely by genetic factors, the existence of the Flynn effect and other cohort influences indicates that the IQ correlations among MZAs and age-matched unrelated pairs are influenced by the birth dates they share.

5. The Cremieux figure omitted the MISTRA control group data. The official MISTRA control group consisted of dizygotic (fraternal) twins reared apart (DZAs). MZAs share 100% of their segregating genes, whereas DZAs share on average only 50%. I showed in a 2022 publication that Bouchard, Segal, and colleagues failed to publish their 1990 DZA IQ correlations (based on three IQ measures). These 1990 control group correlations remain unpublished to this day. Cremieux blocked me on X in 2023 when I pointed to this obvious example of MISTRA p-hacking.

6. The MISTRA MZA and DZA IQ mean difference scores do not differ. The figure I created below is based on the 1990 MISTRA Figure 1 (adapted by Cremieux). In her 2012 book about the study, MISTRA co-investigator Nancy Segal (pp. 286-287) reported the MZA and DZA mean IQ differences at the study’s 2000 endpoint as MZA = 7.07 IQ points based on 70 pairs (standard deviation = 5.83), and DZA = 8.78 IQ points based on 53 pairs (standard deviation = 7.09). These means do not differ at a statistically significant level, and we can safely assume that they didn’t at the 1990 study’s data-collection stop-point either. From the standpoint of inferential statistics, the MISTRA MZA and DZA mean differences are the same. The figure below assumes that the never-published 1990 MISTRA DZA group mean IQ difference was equal to or smaller than that reported by Segal for the larger 2000 sample.

Statistically speaking, the horizontal DZA dashed red line and the solid green MZA line directly below it are the same line. Therefore, because greater MZA genetic similarity did not cause the MISTRA MZA and DZA samples to differ at a statistically significant level, we can conclude that non-genetic influences alone caused smaller MZA and DZA IQ difference scores than would be observed among randomly selected, genetically unrelated pairs matched on age, gender, SES, and other common environmental factors.

The title of the Cremieux figure reads, “The Mean Absolute Difference in IQ between Monozygotic Twins Reared Apart is Minute.” But it is just as “minute” for the 1990 control group DZA pairs. This finding cannot be explained on genetic grounds, but it does help explain why the MISTRA team has kept the 1990 DZA IQ data secret all these years.

For many reasons (fully explained here and here), the MISTRA found no scientifically valid evidence that heredity influences individual differences in IQ scores. The same conclusion holds true for the other studies cited by Cremieux, which did not use a DZA control group.

7. Conclusion. The Cremieux figure provides no valid evidence in favor of genetic influences on differences in people’s IQ scores. That doesn’t mean such influences don’t exist, but because of massive flaws and environmental confounding, TRA studies published to date have failed to uncover them. In many publications (see the list at the end of my 12/13/25 post), I explain in detail how TRA studies, insofar as they are cited in support of genetic influences on human behavioral differences, are a long-running smoke-and-mirrors show. The 2023 Cremieux figure provides a recent example. Pretty colors, artistically arranged. Apart from that, a whole lot of nothing.