Twin Studies and the “Nonreplication Curse” in Psychiatric Molecular Genetic Research

Jay Joseph, Psy.D.

In the spring of 2013, the American Psychiatric Association (APA) officially admitted that it has no genes for its disorders. In an official APA press release dated May 3rd, 2013, DSM-5 Task Force head David Kupfer stated, “In the future, we hope to be able to identify disorders using biological and genetic markers that provide precise diagnoses that can be delivered with complete reliability and validity. Yet this promise, which we have anticipated since the 1970s, remains disappointingly distant. We’ve been telling patients for several decades that we are waiting for biomarkers. We’re still waiting.”[1] In this statement, despite decades of gene discovery claims in professional publications and in the media, and despite the reality that psychiatry has “been telling patients for several decades” that they have faulty genes and chemical imbalances in their brains, American psychiatry came clean about its failure to support its claims with actual scientific findings.

In this context, psychiatry and other fields studying behavior and medical conditions have developed the  “missing heritability” position, which in psychiatry is an attempt to explain the failure to identify genes for psychiatric disorders by claiming that genes exist and await discovery once better methods are found.2

In 2010, researchers developed a new molecular genetic research method, sometimes referred to as “Genome-wide Complex Trait Analysis,” or GCTA. As described by behavioral geneticist Robert Plomin and his colleagues, the GCTA method “compares chance genetic similarity across hundreds of thousands of SNPs for each pair of individuals in a matrix of thousands of unrelated individuals . . . the large sample size makes it possible to estimate heritability directly from DNA markers measured on the microarray.” However, “it does not identify which SNPs are responsible for the heritability of a trait.”3 The method does not identify associated genetic variants such as SNPs (single-nucleotide polymorphism, a common type of genetic variation among people), but rather estimates total genetic variance on the basis of comparing the genetic profiles of a large group of unrelated people. In other words, it can produce a “finding” of genes even when no specific genes are identified.

The original GCTA study, published in 2010 by Yang and colleagues, assessed the genetic basis of human height variation.4 Previous researchers had estimated the heritability of height at 80%, yet genome-wide association (GWA) studies had identified only 5% of the genetic variants responsible. Using the new GCTA method, Yang and colleagues estimated that the proportion of height variance “explained by the SNPs” is 45%.

Although the GCTA method was developed as a means of solving the “missing heritability” problem, like GWA studies and other methods it is based on assuming the validity of heritability estimates for human behavioral traits, as well as assuming that twin and adoption studies have established the genetic basis of psychiatric disorders. The validity of each of these assumptions, however, is very questionable.5 And yet, leading behavioral geneticists such as Robert Plomin and Eric Turkheimer have embraced GCTA, seeing it as a way out of the missing heritability conundrum.6 Turkheimer even saw it as the ultimate refutation of the arguments of twin research critics, writing that “the new paradigm” has put criticism of the assumptions of family and twin studies “to rest.” Turkheimer believed that the GCTA method “should drive a stake through the heart of a classical line of argument against classical behavioral genetics and its attendant statistical assumptions.”7

However, cracks in the GCTA method have already begun to appear. In 2013, Trzaskowski, Plomin and colleagues conducted a study on anxiety-related behaviors in childhood, finding GCTA variance estimates of 1% to 19% (none reaching statistical significance), whereas twin method reared-together MZ-DZ (identical twin pair-fraternal twin pair) comparisons had produced heritability estimates of 50% to 61%.8 In another 2013 publication, Viding, Plomin and their colleagues performed a GCTA study of “callous-unemotional (CU) behavior in children.” Although the authors of previous CU twin studies had estimated heritability ranging from 45% to 67%, the GCTA results showed that “estimates of heritability were near zero.” Viding and colleagues even coined a new term for this discrepancy, calling it the “missing GCTA heritability.”9

Finally, in an October, 2013 study published in the Journal of the American Academy of Child and Adolescent Psychiatry, Trzaskowski, Dale, and Plomin compared GCTA results with twin method results in a study of “childhood behavior problems” which included autistic, depressive, hyperactive, anxiety, and conduct symptoms. The title of their article was clear: “No Genetic Influence for Childhood Behavior Problems from DNA Analysis.”10 The researchers used a large sample of over 2,000 twin pairs, and over 2,000 individuals for the GCTA analysis, and were confronted with a large difference between the twin method results and the GCTA results. The twin study findings reflected the usual behavioral genetic conclusions based on MZ-DZ comparisons, with the researchers calculating heritabilities in the 40%-60% range. The GCTA estimates, however, “are nonsignificant and mostly zero for self-report and parental measures of behavioral problems.” They again referred to this discrepancy as the “missing GCTA heritability.” The researchers spent much of the rest of their publication trying to explain their contradictory findings, describing unexpected results in ways that did not challenge the genetic theories upon which gene searches are based. In doing so, they rejected the possibility that both the GCTA method and the twin method are based on false assumptions and concepts. The most controversial aspect of the twin method has been twin researchers’ theoretical assumption that reared-together MZ and same-sex DZ twin pairs grow up experiencing similar (“equal”) environments. The evidence indicates that this assumption is false.

In an accompanying article, psychiatric geneticist Stephen Faraone, a strong supporter of genetic theories and research in psychiatry, recognized that the 2013 Trzaskowski, Dale, and Plomin study contradicted the findings of the “Cross Disorder Group of the Psychiatric Genomics Consortium” study published in February, 2013.11 Readers may recall that the results of this study were reported with great fanfare in the media as constituting important new gene discoveries in psychiatry (see my March 4th, 2013 MIA blog). As Faraone candidly posed the question, “Nonreplication had been the curse of molecular psychiatric genetics for decades. Has it returned in a new guise?”12

The answer to Faraone’s question probably is yes, because due to major problems and false assumptions underlying previous family, twin, and adoption studies, in addition to the questionable validity and reliability of psychiatric disorders in general,13 the “curse” Faraone referred to is merely the scientific likelihood that genes for the major psychiatric disorders do not exist. This is a cause for celebration, not despair, as society can now part ways with genetic diversions and focus on environmental causes, interventions, and prevention.

Although in 2011 Turkheimer believed that the GCTA method would “drive a stake through the heart of” criticism of behavioral genetic theories and methods, and would finally put criticism of twin studies “to rest,” the opposite scenario appears to be playing out before us, as leading behavioral genetic researchers struggle to prevent negative GCTA findings, and the obvious false assumptions underlying twin research, from driving a stake through the heart of twin studies themselves.

In other GCTA studies showing positive results, it is likely that the method produces false positive findings based on systematic errors and a reliance on questionable assumptions and concepts, similar to those found in GWA studies and other types of molecular genetic research methods that have largely failed to uncover genes for behavioral traits. Evan Charney of the Duke Institute for Brain Science listed several potential biases in GCTA studies, including the failure to adequately account for genetic differences based on variation found among differing populations (population stratification), which introduces a potential environmental confound into GCTA studies. Charney concluded that the GCTA search for thousands of genetic variants of tiny effect “is the last gasp of a failed paradigm.”14 Commenting on a 2011 GCTA study of IQ, critical geneticist Mae-Wan Ho wrote, “This exercise sounds more like a counsel of despair than a solution to the problem.”15

Most likely, these will become the final appraisals of the GCTA method once it runs its course in psychiatry and behavioral genetics, which recent history tells us will be followed by yet another molecular genetic technique promising to deliver the long-lost “missing heritability,” “missing GCTA heritability,”16phantom heritability,” or some other future “missing [fill-in-the-blank] heritability.” Therefore, although the GCTA approach does appear to be a “gasp of a failed paradigm,” it is probably not the last gasp. What actually is “missing” is an independent commission charged with the task of thoroughly examining psychiatric family, twin, and adoption studies (including related theories, assumptions, and original publications) in order to determine whether or not these studies prove anything about genetics.



1 American Psychiatric Association (2013, May 3rd). Chair of DSM-5 task force discusses future of mental health research; Statement by David Kupfer, M.D. American Psychiatric Association [Press release].

2 Joseph, J. (2012). The “missing heritability” of psychiatric disorders: Elusive genes or non-existent genes? Applied Developmental Science, 16, 65-83.

3 Plomin et al. (2013). Behavioral Genetics (6th ed.). New York: Worth Publishers.

4 Yang et al. (2010). Common SNPs explain a large proportion of the heritability for human height. Nature Genetics, 42, 565–569.

5 Joseph, J. (2004). The Gene Illusion: Genetic Research in Psychiatry and Psychology under the Microscope. New York: Algora. (2003 United Kingdom Edition by PCCS Books); Joseph, J. (2006). The Missing Gene: Psychiatry, Heredity, and the Fruitless Search for Genes. New York: Algora; Joseph, J. (2013). The use of the classical twin method in the behavioral sciences: The fallacy continues. Journal of Mind and Behavior, 34, 1-39.

6 Plomin et al. (2013); Turkheimer, E. (2011). Still missing. Research in Human Development, 8, 227-241.

7 Turkheimer (2011).

8 Trzaskowski, Eley, et al. (2013). First genome-wide association study on anxiety-related behaviours in childhood. PLoS ONE, 8(4), 1-7.

9 Viding et al. (2103). Genetics of callous-unemotional behavior in children. PLoS ONE, 8 (7), 1-9.

10 Trzaskowski, Dale, & Plomin (2013). No genetic influence for childhood behavior problems from DNA analysis. Journal of the American Academy of Child and Adolescent Psychiatry, 52, 1048-1056.

11 Cross-Disorder Group of the Psychiatric Genomics Consortium. (2013). Identification of risk loci with shared effects on five major psychiatric disorders: A genome-wide analysis. Lancet, 381, 1371-1379.

12 Faraone, S. (2013). Real progress in molecular psychiatric genetics. Journal of the American Academy of Child and Adolescent Psychiatry, 52,1006-1008.

13 Kirk, Gomory, & Cohen. (2013). Mad Science: Psychiatric Coercion, Diagnosis, and Drugs. New Brunswick, NJ: Transaction.

14 Charney, E. (2013, September 19th). Still chasing ghosts: A new genetic methodology will not find the “missing heritability.” Independent Science News.

15 Ho, M. (2013). No genes for intelligence in the fluid genome. In R. Lerner & J. Benson (Eds.), Advances in Child Development and Behavior, 45, 67-92. San Diego: Elsevier.

16 Zuk et al. (2012). The mystery of missing heritability: Genetic interactions create phantom heritability. PNAS, 109, 1193-1198.

Support MIA

Enjoyed what you just read? Consider a donation to help us continue to produce content, provide up-to-date research news, offer continuing education courses, and continue building a community for exploring alternatives to the current paradigm of mental health. All donations are tax deductible.

Select Payment Method
Personal Info

Credit Card Info
This is a secure SSL encrypted payment.

Donation Total: $20.00


  1. Thank you Jay for your tireless pursuit in laying to rest the psychiatric illusion that human experience can neatly be explained by our genetics.
    Sadly I believe your conclusion is right. Much money and tooting of horns announcing- “this time truly and honestly we are so close to discovering the missing heritability link for ADHD, bipolar, schizophrenia etc, etc, that we can once again act as if it were fact and treat them accordingly” -will continue.

  2. Diseases? Let’s talk about health.!

    What happens when the most VALUABLE, precious genes are found? Great, great worth. Steal the kids?

    What if one’s grandiosity IS genetic? Face the facts, right?

    I’d like to see somebody go further in this direction – identifying genes of exquisite “superiority”. Are they female? Are they Japanese? Divine? Hindu? Blonde? Jewish? Homosexual?

    What great contribution to the whole of Humanity will this ultra-genetic MASTERPIECE serve? The solution to the dreadful plague of the Mind, “mental illness”?

  3. Of course psychiatrists want to blame genes and thus promote eugenics. It seems to be instinctive to the majority of human beings. (Luckily I am a martian.) This is well known in psychology and is called the self-serving attributional bias, and refers to individuals attributing their successes to internal or personal factors but attributing their failures to external or situational factors. The opposite also applies and so arises the tendency to blame other’s failures on their ‘personal’ factors such as genes and brains whilst attributing others success on luck and chance. Of course, now psychiatry is under threat we should expect them to fight back with vigour.

    Thanks Jay for your work on this subject which I have followed with great interest for a long while. Cheers.

  4. “Missing heritability” (meaning “we haven’t found anything”)…What utter garbage. But not a joke to me, because I was locked up, shocked, and tortured my whole childhood because my mom was “schizophrenic,” and so I was said to have inherited the “disease” from her. Very easy to “prove” that, as all they had to do was give me that label.

    Thank you so much for your important work, Jay.

  5. Nice post! It has many interesting and valid points. I believe that the main weakness of all these studies is the diagnostics.

    I should begin by explaining who I am. I am a professor at a research university in North America. I study the statistical aspects of genetics, including inheritance, heritability, and disease.

    I am also a psychiatric survivor with a first-degree relative who is trying to survive. Given my training, you might understand that I was shocked when a psychiatrist conveyed his belief in psychiatric-genetics: since psychiatric diagnoses are genetic, he said, I must have exactly the same diagnosis as my first-degree relative. This statement flies in the face of know principles of genetic uncertainty. Even an animal breeder knows that it is possible to mate a black bull and a black cow to get a red offspring. This belief that psychiatrist have in the inheritance of psychiatric diagnoses borders on religion. It is not science.

    Even before I was diagnosed with a psychiatric disorder, I was involved in the analysis of a psychiatric-genetic study. I was appalled that my psychiatrist colleagues could not make up their minds about which diagnoses to use. Every week for a month, they sent me new diagnostic indicators. Of course I didn’t find anything in the data.

    When one looks at the diagnostics of these genetic studies, one typically finds about 2-5 sentences devoted to describing how the test subjects where diagnosed. The thorough studies will ‘re-diagnose’ the subjects by giving a new clinician access to the subject’s entire health records which undoubtedly include previous diagnoses and diagnoses of relatives. As a result these re-diagnoses largely reinforce the original diagnoses without really providing validation.

    So now, we have studies whose input data is based on the assumption that these diagnoses are inherited, and we try to test whether there is heritability. What do we find? The very heritability that we manufactured in the diagnostic process. This is a case of ‘invented heritability’ as opposed to the ‘missing heritability’ that we find for disease that are actually genetic.

    What do we do about this? Since science is a social process, I think that one good approach is from the inside. Scientists who recognize these flaws need to publish findings that explain the current lack of results. When enough support/disillusion is gathered, scientists sitting on funding panels will grant less funding to poorly designed studies. Perhaps even the funding agencies can be convinced to budget fewer funds for psychiatric-genetic studies.

    So, how do we convince scientists that these studies are flawed? One way would be to document the presence of diagnostic bias that assumes the existence of inheritance. Such a study might invent mental health records for a number of ‘subjects’. The control setting would be that none of these fictitious subjects are related to each other. The setting under investigation would be that there are first-degree relationships between many of these same fictitious subjects. The hypothesis would be that the distribution of diagnoses in the control setting would be much broader and that the diagnoses of related individuals would be unreasonably similar. Here unreasonable would be defined by there being no known genetic factor that could produce such strong similarity. This would be rather strong evidence of the non-genetic basis of these diagnoses.

    Another way to document the influence of such bias is to design a simulation where we have a parameter for the bias. We could then simulate the genomes and diagnoses of related individuals and find our how much of an influence this has on heritability estimates. (This can also be done with real genomes and simulated diagnoses.) Such a simulation would demonstrate that an assumption of inheritance in diagnosing can produce what appears to be evidence of heritability. This would also be a way of modelling how much bias is necessary to produce different degrees of invented heritability.

    I hope to work on some of these things when my other commitments allow it. If anyone has any ideas, feedback, or wants to contribute, please let me know.

  6. It’s great to read real science about the connection of specific genes and specific behaviors; there isn’t any connection. Thank you for challenging the tsunami of pseudoscience supporting the status quo; your work is destined to be part of a classical paradigm shift about the way we understand human behavior.

    Best wishes, Steve

  7. Jay,
    Thank you for working to point out the lack of scientific evidence behind the genetics of the major DSM disorders. Personally, I was made sick with a bad drug cocktail, including a “safe smoking cessation med” / antidepressant and two other drugs (to cover up my PCP’s husband’s “bad fix” on a broken bone). The adverse drug effects were misdiagnosed as “bipolar.” The first neuroleptic resulted in a confessed “Foul up,” according to my medical records. But this “Foul up” resulted in a continuation of the misdiagnosis, and several groups of six or more drugs that all have major drug interaction warnings stating they cause anticholinergic intoxication, a known drug interaction problem that emulates the symptoms of bipolar.

    I don’t want my children or grandchildren to be defamed with a scientifically “lacking in validity disease,” merely because I ran into unethical doctors trying to cover up malpractice by making me sick with known drug interactions. And as an outsider looking in, my research agrees with Whitaker’s, the psychiatric drugs cause the symptoms of the major psychiatric illnesses. The DSM disorders are not genetic illnesses, they are iatrogenic illnesses. Thank you for working to help point out the lack of evidence that they are genetic illnesses.

  8. Excellent highly informative post. Ironically, Vice President Joe Biden’s speech before yesterday’s forum marking the 50th anniversary of the signing of the Community Mental Health Act included the following according to the AP:

    He said science is on the verge of “astounding discoveries” that could change how society cares for those with mental illness. “It’s truly amazing what we don’t know and it’s truly amazing what we might learn,” Biden said during a kickoff of the two-day forum at the Kennedy presidential library. “Imagine when we are able to identify the biomarkers for mental illness.” (Source: Joe Biden Anticipates ‘Remarkable Changes’ For Treatment Of Mental Illness,

  9. Thanks for you superb work, Jay! You are the one of many who offer more evidence the B.R.D. (biological reductionism and determenism) cult is scientifically baseless (and, I think, also philosophically absurd).

    In some decades you (and many other people here on MIA) would be remembered as the ones who stand against the ruling absurdity. And achieved a victory.

  10. Hi, this is Eric Turkheimer. I don’t usually take to time to argue on online forums. But I have to say I get tired of seeing my work mischaracterized by Dr. Joseph. It isn’t that he disagrees with me, that’s fine, and it’s true enough that the GCTA picture has gotten a little more complicated since I wrote the paper he references. But take a minute, read my paper for yourself, and decide whether it is fair to characterize my position as “embracing” GCTA. It is linked here:

  11. Here is a quote from a forthcoming article: Eric Turkheimer, Erik Pettersson, and Erin E. Horn, A PHENOTYPIC NULL HYPOTHESIS FOR THE GENETICS OF PERSONALITY, Annu. Rev. Psychol. 2014. 65:X–X

    The assumptions—both statistical and biological—of the classical twin model have been hotly contended for as long as twin studies have existed, and disagreement about them has not abated (Joseph 2004, Charney 2012). We do not use our limited space debating these issues, for several reasons. They have, of course, been debated many times already. In addition, objections to the assumptions of twin studies are most relevant when the goal of the studies is to compute the heritability of one trait or another, and our explicit goal is to avoid doing so. We have made the case elsewhere (Turkheimer 1998, 2000; Turkheimer & Harden 2013) that the numerical values of heritability coefficients do not matter very much anyway, other than by differing from zero or one. Moreover, some recent DNA-based statistical methods that do not require twins or any assumptions about them have reached conclusions very similar to those from the classical twin studies (Turkheimer 2011, Yang et al. 2011).

    The last sentence is demonstrably false.

  12. I am very pleased that Eric Turkheimer does not embrace the GCTA method – though some of the stuff he’s previously written has certainly given me the impression that he does (“Thanks to the Visscher program…” etc). In distancing himself from it, however, he ought to realise that the old behavior genetic (twin etc) methods and the GCTA method are based on common presuppositions that are genetically absurd. Imagine upholding the claim that at least half of all genes associated with complex psychological traits are utilised randomly and independently of one another and the environment! Surely any results that claim to validate that view simply depicts the triumph of flawed methods and increasingly opaque statistics over the real molecular biology of development. A new paradigm is now beginning to eclipse the old and it will not allow keeping a foot in both camps.

  13. Well I don’t want to argue too much here. Dr. Joseph didn’t answer my question: on a careful reading of the original paper he cited (not the one e quoted above), would he really describe it as embracing GCTA? Take my word for it, I don’t. And here is the thing about my work: I neither “embrace” GCTA or reject it. GCTA is what it is, all there is to do is to think about it, consider what it means and what it doesn’t mean. So I disagree with the commenter above who says you can’t have your feet in both camps. You can, I do. A final word is that, all talk of behavior genetics aside, I have always had a great deal of sympathy with the kinds of ideas expressed here about mainstream psychiatry. It’s just that I don’t think those kinds of ideas are really at odds with thoughtful consideration of the role of genetics in behavior, if it is understood properly. Anyway, Peace. Eric

  14. Jay, yes, I think more dialog would be a good thing. But, as has been the case in some “negotiations” on the national level, for it really to be a negotiation both sides have to give a little. I have been known as a critic of mainstream BG for twenty years. Is there any part of your work that concedes that twin studies can accomplish certain limited goals? Here, for example, is my most recent discussion of the methodological basis of BG. It is conducted entirely without reference to heritabilities or variance computing, except to criticize them. Yet it finds an important role for twin studies in social science.

    By the way, your statement about “failure to uncover genes for behavioral traits and disorders” is an oversimplification. As sample sizes have gotten bigger and bigger, the gene finders are identifying SNPs correlated with, say, schizophrenia at more and more stringent levels of significance. So for me, the question is not whether there are any correlations there, but whether they add up to anything interesting. The difference between you and me, I think, is that you want to reject outright any correlation between phenotype and genotype, so those SNP correlations had better be zero. I say that very general relations between genotype and phenotype are a fact of how organisms are put together, so ultimately “successful” but totally spread out and probably theoretically uninteresting SNP findings are a consequence of that organization.


  15. This post is brilliant.

    The idea of “meeting mid-way” with a failed paradigm that cannot deliver the goods makes as much sense as “meeting mid way” in arithmetic.

    2 + 2 equals 4. You cannot meet mid way with somebody who keeps on insisting that 2 +2 equals 7…or maybe 10…0r maybe whatever!