Chapter 5

Schizophrenia Adoption Studies

“The turning point in schizophrenia research, and probably the most important data collection and results of the 20^th century in this field, came from the carefully planned and executed adoption studies of Seymour Kety and David Rosenthal….These data…turned the corner for support for ‘nurture’ to support for primarily ‘nature.’”

—Schizophrenia genetic researcher Lynn DeLisi, in her 2017 book 100 Questions & Answers About Schizophrenia: Painful Minds²¹⁵

Because family and twin studies have been rightly criticized since their inception as being unable to disentangle (separate) the potential influences of genes and environment, schizophrenia adoption studies were performed in an attempt to finally accomplish a clean separation between these potential influences. Schizophrenia adoption researchers believed that their studies were able to accomplish this because adoptees inherit the genes of their biological (birth) parents, but are reared in the environment of another (adoptive) family with whom they share no genetic relationship. From Rosenthal’s perspective, unlike family and twin studies, adoption studies are able to “separate the genetic and rearing variables entirely.”²¹⁶ Based on this belief, leading psychiatric genetic researchers have written, “By far the most convincing evidence for a genetic contribution to schizophrenia is that provided by a series of adoption studies.”²¹⁷ However, we will see that there were numerous problem areas in these studies, and like family and twin studies, it is unlikely that they were able to adequately separate the potential effects of genes and environments.²¹⁸

Complete Separation of Genes and Environment?

Abandoned Children

We must always remember that the “adoptees” in these studies were children abandoned by, or taken from, their birthparent(s) for various reasons, often under difficult conditions in early-to-mid 20^th century Europe and the United States. It is likely that most experienced attachment-rupture trauma, emotional suffering, loneliness and neglect, abuse, and other types of hardships. This was especially true for the late-separated children, and for children who spent time in an orphanage. As a group of French critics wrote, a more fitting name for this area of research would be the study of abandoned children.²¹⁹ This evokes a different set of emotional responses than the more positive, trauma-obscuring term “adopted children” (trauma-obscuring in the sense that it places emphasis on the fact that children were placed into adoptive homes, not on how they got there).

Research performed since the 1990s has shown that disturbed or ruptured parent-child attachment patterns can influence brain development during critical developmental periods.²²⁰ As the authors of a 2015 review put it,

“Environmental influences in infancy, particularly the quality of the caregiver–infant relationship and emotional interactions within this context, have been purported to shape neurological, psychological and social development and have potential long-term effects on psychological and emotional functioning.…Early deprivation of comfort and security has been found to have adverse sequelae on a broad range of domains, including neurological, psychological, emotional and physical development and functioning.”²²¹

These findings are consistent with Read and colleagues’ Traumagenic Neurodevelopmental model of psychosis, and call into question adoption researchers’ claim that their conclusions apply (generalize) to the non-adoptee population. Later we will see that many abandoned children whose records were used in schizophrenia adoption research were reared by their birthparent(s) for several months or even years, and languished in orphanages and foster homes between separation and adoption.

Prenatal Environment

In his 2010 book on addiction, In the Realm of Hungry Ghosts: Close Encounters with Addiction, physician Gabor Maté emphasized the importance of a child’s prenatal environment, which takes place in the body of his or her stressed (and possibly distressed) birthmother, as a problem area in adoption research:

“Any woman who has to give up her baby for adoption is, by definition, a stressed woman. She is stressed not just because she knows she’ll be separated from the baby but primarily because if she wasn’t stressed in the first place, she would never have had to consider giving up her child: the pregnancy was unwanted or the mother was poor, single, or in a bad relationship; or she was an immature teenager who conceived involuntarily or was a drug user or was raped or was confronted by some other adversity.”²²²

Maté noted that “any of these situations would be enough to impose tremendous stress on a person,” meaning that “for many months the developing fetus would be exposed to high cortisol levels through the placenta,” which can affect the developing brain and nervous system. He also discussed the “significant impairment” suffered by infants developing in the distressed post-natal rearing environment of a woman whose life was so chaotic that she was compelled to give her child up for adoption.²²³ In a later communication Maté pointed out that, theoretically speaking, to avoid these important problems “adopted children would need to be randomly selected from ‘well’ families who had no intention of giving up their children.”²²⁴

At the very least, a valid schizophrenia adoption study would need to randomly select newborn children with a biological family history of schizophrenia, and then randomly assign them at birth to the full range of available adoptive homes. Prenatal environments would be assessed as much as possible.²²⁵ Standardized diagnoses would be made decades later by neutral qualified researchers blind to the group status of the adoptees and relatives. The experimental and control group adoptees, and their adoptive families, would have no knowledge of the psychiatric status of the adoptee’s biological family. The researchers’ intended methods and diagnostic criteria would be written up and “preregistered” before the work begins (see below). Researchers would also be required to provide full access to their raw data to qualified reviewers and scientists.

Range Restriction and Representativeness

The restricted socioeconomic range of adoptive family environments is another non-genetic factor influencing the behavioral resemblance of adoptees and their adoptive parents.²²⁶ Parents who adopt children are not representative of the full range of potential adoptive parents for other reasons. As Rosenthal acknowledged, “The screening with respect to adopting parents is well known, since adoption agencies have long taken the view that mentally ill people do not make the kinds of parents that serve the best interests of the child.”²²⁷

Birthparents who give up a child for adoption also are not representative of the population of birthparents. Kendler and his psychiatric genetics colleague Pak Sham discussed in 2008 “several potential methodological limitations” in psychiatric adoption research, one of which is “unrepresentativeness.”²²⁸ This refers to the fact that “biological parents who give children up for adoption…have higher rates of psychiatric illness and drug problems (although several generations ago, a more likely problem was poverty).” In contrast, “adoptive parents are typically mentally healthier than the general populace because adoption agencies see it as their job to select ‘healthy’ families in which to place their adoptees.” Plomin and colleagues also pointed to these potential problems: “If biological parents, adoptive parents, or adoptive children are not representative of the rest of the population, the generalizability of adoption results could be affected.”²²⁹

Because adoptive parents are screened for mental health problems, it is not surprising that studies find fewer psychiatric disorders among them. Obviously, comparing diagnostic rates among adoptive parents screened for mental health problems, versus birthparents having “higher rates of psychiatric illness and drug problems,” has little meaning as it relates to genetic research. Nevertheless, one of the two adoption studies of bipolar disorder (manic-depression) frequently cited in favor of genetics was based on this comparison.²³⁰

Sham and Kendler believed that due to the “methodological problems” of twin and adoption studies, “convincing evidence for a genetic component should preferably come from both types of study design.”²³¹ But if both methods are subject to environmental confounds and other biases, combining what may be genetically misinterpreted results in each study still does not add up a genetic finding. This “converging evidence” argument has long been used by proponents of questionable scientific claims who typically maintain, as one group of commentators wrote, “that scientific claims can be evaluated only within the context of broader claims and therefore cannot be judged in isolation.” This argument allows supporters of dubious or unscientific techniques, methods, or belief systems to “readily avoid subjecting their claims to the risk of falsification” by directing attention elsewhere in support of their claims.²³²

***

It is often stated in the genetics literature that twinning, and the adoption process, provide excellent “natural experiments” for separating the effects of genes and environment (nature and nurture). We have seen that this claim is wrong as it relates to twin studies, and adoption studies are also subject to the confounding influence of environmental factors. An additional important issue is the fact that, in the countries and eras in which schizophrenia adoption research was performed, matching between biological and adoptive families was influenced by the perceived genetic background of the child. I discuss adoption studies’ “selective placement” problem later in this chapter.

Bias and Deception in Behavioral Research

Like other areas of research, schizophrenia adoption researchers arrived at conclusions in favor of genetics after producing results that fell below the conventional .05 level of statistical significance. This means that there was less than a 5% probability that the genetic finding occurred by chance. Above the .05 level, they would have been required by convention to conclude that they had found no evidence in support of genetic influences on schizophrenia. Larger sample sizes increase the likelihood that differences between groups will reach statistical significance; smaller samples have the opposite effect. In the adoption studies that I am about to review, a probability value (“p-value”) below the .05 threshold was the researchers’ make-or-break gold standard, enabling them to conclude in favor of genetics.

In normal experiments based on the standard “hypothetico-deductive” (H-D) scientific method, in sequence researchers generate hypotheses, design a study, collect data, analyze data and test hypotheses, interpret data and determine statistical significance, and submit their findings for publication. In the process they perform “null hypothesis significance testing” (NHST). The “null hypothesis” is a default position which states that there is no difference between the specified populations under study, and that any observed differences are due to chance, or to experimental error. In schizophrenia adoption research, the null hypothesis states that there is no difference between the schizophrenia experimental group versus the control group, meaning that the study found no evidence that genetic factors play a role in causing the condition. If researchers find group comparisons below the .05 threshold in the genetic direction, they reject the null hypothesis and conclude that hereditary factors are responsible for the group differences. Researchers in all fields are expected to formulate their hypotheses before they collect their data.

In 2017, psychologist Chris Chambers published The Seven Deadly Sins of Psychology: A Manifesto for Reforming the Culture of Scientific Practice.²³³ Chambers pointed to several problem areas in the research/publication process in psychology and other fields. These include the “deadly sins” of “bias,” “hidden flexibility,” “unreliability,” “data hoarding,” “corruptibility” (fraud), and “bean counting” (funding and publication issues). There are many other problem areas in social and behavioral science research that were not covered in this book, yet those Chambers highlighted are very relevant to the body of schizophrenia adoption research that we are about to examine.

P-Hacking, HARKing, and Data Dredging

P-Hacking. Although many of the “sins” Chambers discussed are found in behavioral genetic and psychiatric genetic research, p-hacking and HARKing are most relevant to schizophrenia adoption research. P-hacking is the practice of consciously or unconsciously manipulating data to produce results that fall below the .05 level of statistical significance. Researchers have “degrees of freedom” that allow them the “hidden flexibility” to change various aspects of their study after reviewing the data, but before submitting their paper for publication and peer review. As Chambers defined it, p-hacking is “exploiting researcher degrees of freedom to generate statistical significance.” A “key feature” of researchers’ decisions “is that they are hidden and never published.”²³⁴ P-hacking occurs, as a group assessing its impact put it, “when researchers collect or select data or statistical analyses until nonsignificant results become significant.”²³⁵

Some ways that researchers can p-hack data include (1) conducting analyses midway through experiments to decide whether to continue collecting data (“peeking” at data), and stopping data exploration if an analysis yields a statistically significant p-value; (2) recording many response variables and deciding which to report after-the-fact; (3) deciding after-the-fact whether to include or remove outliers; (4) excluding, combining, or splitting treatment groups after-the-fact; and (5) continuing to collect data past the planned stop point if significant comparisons are not found.²³⁶ Because social and behavioral science researchers have the hidden flexibility to change definitions and methods without else anyone knowing, as Chambers noted they are able to decide when to stop counting participants (subjects), and are able to redefine the “dependent measure” (schizophrenia in the present context). This enables researchers to “navigate either deliberately or unconsciously in order to generate statistically significant effects.”²³⁷ Surveys suggest that this type of behavior is common in psychology (and presumably in other social and behavioral sciences as well), and occurs in part because there are many built-in incentives and pressures in academic research to p-hack, and few safeguards in place to prevent it.²³⁸

HARKing. The term “HARKing” was introduced by psychologist Norbert Kerr in 1998, and stands for “hypothesizing after the results are known.”²³⁹ Kerr defined HARKing “as presenting a post hoc hypothesis in the introduction of a research report as if it were an a priori hypothesis.”²⁴⁰ In other words HARKing occurs when, after researchers inspect their data, they create a new hypothesis that they claim or imply was created before they inspected their data. In Chambers’ words, “HARKing is a form of academic deception in which the experimental hypothesis (H1) of a study is altered after analyzing the data in order to pretend that the authors predicted results that, in reality, were unexpected.” This method produces the “clean and confirmatory papers that psychology journals prefer while also maintaining the illusion that the research is hypothesis driven and thus consistent with the H-D method.” Chambers concluded that “deliberate HARKing…lie[s] on the same continuum of malpractice as research fraud.”²⁴¹ Again, there are few safeguards in place to prevent HARKing. The peer-review process in science, which usually takes place after a paper is submitted for publication, is not equipped to detect HARKing or p-hacking, even if peer reviewers wish to do so.

Data Dredging. Another unsound research practice is “data dredging” (also known as a “fishing expedition”), which involves searching through data in an attempt to find statistically significant trends or differences, without testing a prior hypothesis. Identifying correlations and potential factors can be useful to help arrive at a hypothesis, but that hypothesis must then be tested on a different set of data. As the authors of a medical textbook emphasized, a hypothesis cannot be developed and tested in the same study. If this happens, data dredging has occurred:

“The scientific process requires that hypothesis development and hypothesis testing be based on different data sets. One data set is used to develop the hypothesis or model, which is used to make predictions, which are then tested on a new data set” (italics in original).²⁴²

Data dredging is related to the “Texas sharpshooter’s fallacy,” which describes a sharpshooter who fires his gun at the side of a barn, and later draws targets around a cluster of points that were hit. People viewing the barn might think that he meant to hit these targets, which were not drawn until after the gun was fired. According to Wikipedia, this “fallacy is characterized by a lack of a specific hypothesis prior to the gathering of data, or the formulation of a hypothesis only after data have already been gathered and examined.” It is a fallacy in part because, in a large dataset based on multiple comparisons, we would expect to find statistically significant correlations by chance alone.

Although data dredging is a form of p-hacking, researchers can select statistically significant results or comparisons after the fact without manipulating their data to do so. Data dredging also differs slightly from HARKing because, although researchers are pointing to comparisons that they did not plan to highlight, they are not necessarily claiming that they are testing a prior hypothesis.

The Preregistration of Research

P-hacking, HARKing, and data dredging are methods that some researchers use consciously or unconsciously to achieve statistically significant results, even though the null hypothesis may in fact be true. There are a number of possible motivations for doing this. Scientific researchers are under pressure to produce statistically significant findings in order to get their studies published, which might tempt them to use their “degrees of freedom” to produce results that journals will publish. Other possible motivations include financial, a desire to achieve career advancement and prestige, supporting their field against critics, and ideological. Genetic (biological) determinism is an ideology, although its adherents usually deny this and claim that their beliefs are based on nothing more than objective scientific evidence.

Building on calls by previous authors going back to the 1960s, which includes my own 2000 proposal co-authored by psychologist Steve Baldwin, Chambers called for the establishment of psychology research “preregistration,” where investigators would be required to submit an introduction, and their proposed methods and analyses, before they collect their data.²⁴³ As Chambers described it,

“The essence of preregistration is that the study rationale, hypotheses, experimental methods, and analysis plan are stated publically in advance of collecting data….Since authors will have stated their hypotheses in advance, preregistration prevents HARKing and ensures adherence to the H-D [normal] model of the scientific method…Preregistration also prevents researchers from cherry-picking results that they believe generate a desirable narrative.”²⁴⁴

The preregistration of research would greatly reduce p-hacking, HARKing, data dredging, and other deceptive methods. Fortunately, a movement is now underway to make preregistration the norm in the social and behavioral sciences. In the upcoming analysis of the Danish-American adoption studies, I will point to several instances where the researchers clearly or likely resorted to p-hacking, HARKing, or data dredging. Of course, research preregistration did not exist during the time these studies were carried out.

The Individual Studies

Psychiatric genetic researcher Elliot Gershon and his colleagues summarized the impact of the Oregon and Danish-American schizophrenia adoption studies in supporting genetic theories, writing in 1994, “The adoption studies of schizophrenia by Heston and by Kety et al. had the seminal effect of establishing that a transmissible cause of schizophrenia predated the adoptions….This ruled out the prevailing causative theories that attributed the illness to mother-child interaction at a later age.” They noted that “these studies turned a historical tide in psychiatric thought and lent credibility to the decades of twin and family studies.”²⁴⁵ These studies also played a major role in reviving the psychiatric genetics field from its post-World War II decline, with Gottesman and Shields believing that they provided “the straw that broke the environmentalist’s back.”²⁴⁶

The Danish-American adoption studies, performed in the 1960s-1990s by American psychiatric genetic researchers Seymour Kety, David Rosenthal, Paul Wender, and their Danish colleagues which included Fini Schulsinger (1923-2012) and Joseph Welner, are the most frequently cited investigations, and are therefore the main focus of this chapter. Seymour Kety (1915-2000) was a highly regarded psychiatric researcher who had developed the first quantitative technique for measuring blood flow in the living brain. Kety first proposed the idea of performing an adoption study of schizophrenia in a 1959 article published in Science.²⁴⁷ David Rosenthal (who died in 1996) was a psychologist at the U.S. National Institute of Mental Health. He was a supporter of the German psychiatric genetics tradition of Rüdin and his associates, and was the editor of The Genain Quadruplets: A Study of Heredity and Environment in Schizophrenia.²⁴⁸ This 1963 book described a set of reared-together identical female quadruplets, all of whom were diagnosed with schizophrenia. Rosenthal’s commitment to genetic theories, predating the first Danish adoption study publications, led him to minimize the impact of the extreme abuse and isolation he described these young women as having suffered growing up.²⁴⁹ Paul Wender (1934-2016) was a psychiatrist who later was involved in ADHD research. In 1962 Kety, Rosenthal, and Wender decided to pool their efforts, and with Schulsinger’s help they began the study in Denmark in 1963.²⁵⁰

The story that mainstream sources fail to tell, however, is that their studies’ glaring and massive flaws, biases, denial or minimization of the impact of environmental confounds, and arbitrary after-the-fact juggling of diagnoses helped ensure that the researchers would arrive at conclusions in favor of genetics. To find these glaring flaws and biases it not necessary to comb the archives, or to uncover unpublished documents and smoking guns. They are found, for the most part, by reading the investigators’ publications closely, and by understanding the history and methods of psychiatric genetic research as well as the motivations and biases of the original investigators and their backers.

In the first five adoption studies of schizophrenia, the researchers compared the diagnostic rate of their schizophrenia “index” group (experimental group) adoptees or biological relatives, versus the diagnostic rate of their control group adoptees or biological relatives. In the sixth, Lichtenstein and colleagues did not use an adoption control group. The Kety-led studies began with adoptees as the first-identified relatives or probands (see Figure 2). The Heston, Rosenthal-led, and Tienari-led studies began with diagnosed biological parents as the first-identified relatives or probands (see Figure 3). In addition to parent/adopted-away offspring pairs, the Lichtenstein-led study calculated sibling/adopted-away sibling risks.

Heston’s U.S. Oregon Study

In 1966, genetically oriented psychiatric researcher Leonard Heston assessed the rate of schizophrenia among the 47 adopted-away biological offspring of women diagnosed with schizophrenia, who were confined to Oregon state mental hospitals.²⁵¹ Heston, who was not blind to the group status of the adoptees, along with two colleagues made 5 schizophrenia diagnoses among these 47 experimental group adoptees, versus zero among the 50 control adoptees, a result that just reached statistical significance (p = .024). Heston concluded that his results “strongly support a genetic aetiology of schizophrenia.”²⁵²

In addition to the general problems of psychiatric adoption research, selected problems specific to Heston’s study include (1) that schizophrenia was not defined (Heston stated only that schizophrenia diagnoses were made “conservatively,” and were based on “generally accepted standards”²⁵³); (2) a diagnostic process that was contaminated by a non-blinded rater; (3) that almost no case history material was provided, which would have allowed independent analysis of adoptees’ history and mental/diagnostic status; (4) that about 25% of the adoptees were not interviewed, yet were retained in the study; (5) that almost half of the adoptees had spent months or years in an orphanage, with Heston admitting that “none of the subjects [was] reared in typical or ‘normal’ circumstances”²⁵⁴; and (6) that the results were biased by selective placement factors (see below). A more detailed critical analysis of Heston’s study can be found elsewhere.²⁵⁵

The Kety-Led Danish-American Studies

The most-frequently cited of the three major Danish-American adoption studies is Kety, Rosenthal, Wender and colleagues’ “Adoptees’ Family Study,” carried out in two geographical regions of Denmark beginning in 1963. The first publication appeared in 1968, and the final publication appeared in 1994. In each of these two studies, the preliminary report was based on diagnoses obtained from institutional records, while in the final report the researchers made diagnoses on the basis of interviewing available relatives.

The 1968 Copenhagen Records-Based Study. The original 1968 “Copenhagen Study” began with the records of 5,483 Danish children adopted by non-relatives in the city and county of Copenhagen between 1924 and 1947.²⁵⁶ Of these adopted (abandoned) children, records from Danish psychiatric registers identified 507 who later in life had been admitted to a Danish psychiatric facility for any reason. Based on these records, the researchers diagnosed 33 adoptees with “chronic,” “acute,” or “borderline (latent)” schizophrenia. These and other “schizophrenia spectrum diagnoses” will be discussed in more detail below.²⁵⁷ These became the 33 “index adoptees.” The researchers also established a control group, which consisted of 33 adoptees with no record of admission to a Danish psychiatric facility. Each control adoptee was matched to an index adoptee on the basis of age, sex, socioeconomic status of the rearing family, and time spent with biological relatives, in a child-care institution, or in a foster home before transfer to the adopting family. Kety and colleagues then made blind, consensus diagnoses among the 306 identified first- and second-degree biological relatives of these index and control adoptees (150 index, 156 control). They performed the same procedure for the 157 index and control adoptive (rearing) relatives. The four groups of relatives are seen in Figure 2. Based on finding a statistically significant schizophrenia spectrum disorder rate in the index biological relative group compared with the control biological relative group, Kety and colleagues concluded that “genetic factors are important in the transmission of schizophrenia.”²⁵⁸

The 1975 Copenhagen Interview-Based Study. In a widely cited 1975 follow-up study, the researchers assessed the diagnostic status of the 1968 Copenhagen biological and adoptive relatives, but here they did so through the use of psychiatric interviews.²⁵⁹ These were the first- and second-degree relatives of the 33 index and 33 control adoptees identified in the 1968 study, although the control group now contained 34 adoptees.²⁶⁰ The biological relative group had now grown to 347 (173 index, 174 control). The researchers undertook the difficult task of tracking down and attempting to interview these relatives. They believed that there were many more “schizophrenia-related disorders” among them that had not come to the attention of mental health facilities or hospitals, and were therefore unrecorded in the 1968 study. Due to death, unavailability, or refusal, the researchers were able to interview 72% of the identified biological relatives, and 48% of the identified adoptive relatives.²⁶¹ The interviews were performed by the Danish members of the research team, and were written up in English as a 35-page transcript. These transcripts were edited to remove information that could reveal the relative’s group status, and were sent to Kety, Wender, and Rosenthal in the U.S., who made blind consensus psychiatric diagnoses. The code was then broken, and the relatives (and their possible diagnoses) were allocated to their respective groups as defined by their relationship to the index and control adoptees. Based on finding a significantly higher rate of schizophrenia spectrum diagnoses among the index biological versus the control biological relative groups, in addition to other comparisons, the researchers concluded that their results were “strongly suggestive of the operation of genetic factors.”²⁶²

The Provincial Study. After the completion of the Copenhagen 1968 records-based and 1975 interview-based studies, the investigation was extended to the remaining provinces of Denmark. The preliminary records-based results of this “Provincial Study” were published in 1978. The final interview-based Provincial Study results were not published until 1994. The researchers combined the Copenhagen and Provincial results to produce the “Danish National Sample.”²⁶³ Kety and colleagues concluded:

“This study and its confirmation of previous results in the Copenhagen Study speak for a syndrome that can be reliably recognized in which genetic factors play a significant etiologic role. These findings provide important and necessary support for the assumption often made in family studies: observed familial clustering in schizophrenia is an expression of shared genetic factors.”²⁶⁴

A year prior to his death in 2000, Kety won the prestigious Lasker Special Achievement Award in Medical Science for his adoption study work, and for his earlier studies of cerebral blood flow.

The Rosenthal-Led Danish-American Study

Unlike the Kety-led studies, which began with adoptees as the first identified relative (proband), in Rosenthal, Kety, Wender and colleagues’ “Adoptees Study” the first-identified relatives were biological parents who had given up a child for adoption. The preliminary publication appeared in 1968. A follow-up study, based on an enlarged sample, was published in 1971.²⁶⁵ The researchers began with the records of the same 5,483 Danish adoptees identified in the Kety-led study, and then identified around 10,000 of their biological parents.

Based on institutional records, Rosenthal, Kety, and Wender identified 76 Danish biological parents (50 mothers, 26 fathers) who they diagnosed with “schizophrenia,” “doubtful schizophrenia,” or “manic-depressive psychosis.”²⁶⁶ The 76 adopted-away (abandoned) biological offspring of these parents constituted the index adoptee group. The researchers compared the diagnostic rate found in this group to a matched control group consisting of the 67 adopted-away offspring of parents with no known psychiatric history. The basic design, also used in the Heston and Tienari-led studies, is seen in Figure 3.

In lieu of a consensus diagnosis of each adoptee, in the 1968 and 1971 studies the researchers used what they called “thumbnail diagnostic formulations.”²⁶⁷ All 1968 and 1971 diagnoses, statistical comparisons, and conclusions in favor of genetics were based on these thumbnail formulations (or impressions), which were prepared by Welner and Schulsinger on the basis of a 3-5 hour interview with each adoptee. Some examples of such diagnostic formulations/impressions in the 1971 study included “possible paranoid borderline,” “paranoid character,” “almost pseudoneurotic borderline,” “moderately schizoid,” “pronounced preschizophrenic diathesis,” and “conceivably paranoid borderline.” In their statistical calculations, Rosenthal and colleagues counted these formulations as schizophrenia spectrum disorders, although they did not state who made this determination. They calculated a 31.6% (24/76) schizophrenia spectrum disorder rate among the index adoptees, versus a 17.8% (12/67) rate among the control adoptees. This comparison reached statistical significance just below the .05 level, allowing Rosenthal and colleagues to conclude, “The authors believe that evidence from this study supports the theory that heredity plays a significant role in the etiology of schizophrenia spectrum disorders.”²⁶⁸

The Wender-Led Danish-American Study

In the least-frequently cited Danish-American study, the 1974 “Crossfostering Study” by Wender, Kety, Rosenthal and colleagues, the researchers investigated the adopted-away biological offspring of Danish parents not diagnosed with schizophrenia, but who were raised by an adoptive parent eventually diagnosed with a schizophrenia spectrum disorder.²⁶⁹ The stated purpose of the Crossfostering Study was based on the following question: “Can psychopathology in rearing parents produce psychopathology in the offspring when the offspring do not carry a genetic load for schizophrenia?”²⁷⁰ Wender and colleagues believed that their study provided

“a technique for clarifying the role of experiential factors in psychopathology—the crossfostering technique—employs the strategy of using adopted-away offspring of normal biological parents who are reared by schizophrenic parents. This group is compared with the adopted-away offspring of normal biological parents reared by normal adopting parents and the adopted-away offspring of schizophrenic biological parents reared by normal adopting parents.”²⁷¹

The researchers’ “tentative conclusion” was that their study showed “that genetic factors play a role in the etiology of the schizophrenias while familial psychopathology (as measured by psychiatric diagnosis of the parents) does not.”²⁷²

The Tienari-Led Finnish Study

In contrast to the earlier investigations, Pekka Tienari and colleagues’ 1980s-2000s “Finnish Adoptive Family Study of Schizophrenia” took the important step of looking at adoptive family environments as well as adoptees’ genetic background.²⁷³ Their index group consisted of the 190 adopted-away biological offspring of mothers diagnosed with DSM-III-R schizophrenia and other “schizophrenia spectrum disorders.” Tienari and colleagues reported a statistically significant index versus control difference for the “schizophrenia spectrum” disorders combined. For “typical schizophrenia” alone, however, there was no statistically significant index versus control group difference.²⁷⁴ They concluded that in addition to genetic background, adoptive family environments “were a significant predictor of schizophrenia spectrum disorders in adoptees.”²⁷⁵

In addition to the general problems of psychiatric adoption research, selected problems specific to Tienari’s study include (1) the above-mentioned failure to find a statistically significant index versus control schizophrenia difference, (2) that about one-third of the adoptees were placed after their mother was diagnosed with a psychotic disorder, (3) that the definition of schizophrenia was broadened to include “schizophrenia spectrum disorders,” (4) that some children were adopted away as late as age four, and (5) that the results were biased by selective placement factors (see below).²⁷⁶ Detailed critical analyses of the Tienari-led study are available elsewhere.²⁷⁷

The Lichtenstein-Led Swedish Study

In 2009, Paul Lichtenstein and colleagues performed a large study of relative risks in Sweden, based on records obtained from a population register and a hospital discharge register. Their main purpose was to assess whether schizophrenia and bipolar disorder “share common genetic determinants.” When studying adoptive relationships, they found an elevated ICD diagnostic criteria schizophrenia risk among the adopted-away biological offspring of parents diagnosed with schizophrenia, as well as among the adopted-away siblings of non-adopted siblings. Lichtenstein and colleagues did not use an adoption control group. They did not state how many adoptees they studied, or provide information relating to adoptees’ placement or age at separation.²⁷⁸

Major Problem Areas of the Danish-American Adoption Studies

In all schizophrenia adoption studies the investigators concluded that genetic factors play a role in causing the condition, while Tienari concluded that disturbed family environments also play such a role. Although in general the Danish-American researchers recognized “that environmental factors are necessary for the development of schizophrenia,”²⁷⁹ in a 1977 article Wender, Rosenthal and others wrote, “With the exception of one finding in one experiment, adoption studies have demonstrated that genetic and not rearing factors play an etiological role in the schizophrenias.”²⁸⁰ The researchers, however, provided very little case history information for adoptees or relatives, and failed to assess the role of key environmental factors in the adoptive (rearing) families.

The Danish American adoption studies have been the subject of several major critical reviews.²⁸¹ They were also the focus of several “independent” analyses, performed by genetically oriented psychiatrists and psychologists in the 1980s and 1990s, who blindly reanalyzed the original Danish-American adoption study data on the basis of DSM-III diagnostic criteria. These analyses (or “reanalyses”), which were published in leading psychiatry journals, are frequently cited in textbooks as having “reaffirmed” or “upheld” the original Danish-American findings. I will refer to these analyses only in passing, since my primary objective is to show that the original schizophrenia adoption studies themselves produced no genetic findings to “reaffirm.”²⁸²

Unlike DSM-III (1980) through DSM-5 (2013) diagnostic criteria based on the more objective (yet still very problematic) method of checking off a list of symptoms, the Danish-American researchers diagnosed adoptees and relatives using the older “global” (or “consensus”) diagnostic system based on DSM-II descriptions (1968).²⁸³ They described this method, which they used in the entire series spanning more than 25 years, as follows:

“Four copies of the edited summary were prepared and distributed to the four authors who served as raters and who independently characterized each subject according to the classification described below [see Figure 4]. The individual ratings were then tabulated and those cases in which there was disagreement among the raters were discussed at a conference of all four authors where an effort was made to review additional edited information which it was possible to obtain and to arrive at a consensus diagnosis acceptable to all. In 4 cases there remained an evenly split opinion regarding the presence of schizophrenia or doubtful schizophrenia, and these were not included in those categories.”²⁸⁴

Although Kety claimed “a high degree” of diagnostic reliability in his studies, each investigator approached the diagnostic process somewhat differently.²⁸⁵ According to Kety, each rater’s “individual definitions of schizophrenia varied by virtue of [their] training and experience, from a substantial reliance on Kraepelin and Bleuler to the broader psychodynamic concepts which were taught in the 50s.”²⁸⁶ This was the era in which, as we saw earlier, Penrose concluded that the “study of the genetics of schizophrenia is unsatisfactory from almost every point of view,” because “there is no certainty that the condition can be defined or even recognized.” We saw that the authors of the “Cross-National Project,” also carried out in this era, concluded that schizophrenia “diagnoses routinely made in clinical practice should not be relied upon in epidemiological studies.”

At the close of the June, 1967 conference in Dorado, Puerto Rico that saw the Danish-American investigators’ first public presentation of their results (later published in the 1968 book The Transmission of Schizophrenia), Rosenthal noted that environmentally oriented researchers like to focus on people (“patients”), whereas genetically oriented researchers prefer to focus on numbers and statistics.²⁸⁷ The Danish-American researchers epitomized the “numbers and statistics” approach. The compelling human experience of psychosis, and the relationships, environments, oppression, and other social contexts that help produce it, are largely absent from their publications. In their place are numbers, diagnoses, tables, statistical tests of significance, population and psychiatric register data, and degrees of genetic relationship—all presented in a not-so-neat package that nonetheless led to the conclusions that psychiatry and other vested interests needed.

I use the word “decided” frequently in this chapter to emphasize the point that scientific researchers are faced with numerous decisions about what to study, how to define variables, whether to publish, where to publish, what comparisons to make, and so on. The public is only presented with what they decide to publish, and Kety, Rosenthal, and Wender often had the results at hand when they made various decisions. The word “decided” or “chose” could be italicized frequently in Chapter 5, because a different set of decisions on the part of the Danish-American researchers would have led to a completely different set of conclusions.

I will now discuss several major problem areas in the Danish-American adoption studies, although by no means are these the only problem areas. Most of the following points have been made by critics over the years, and I will elaborate on these points and add several new ones. Because these studies have played a major role in supporting genetic and biological theories in psychiatry and other fields, it is necessary to examine these problem areas in detail.

The Schizophrenia Spectrum Concept

We have seen that the Danish-American investigators decided to expand the definition of schizophrenia to include what they called schizophrenia spectrum disorders, and they would not have found statistically significant results without such an expansion. Danish-American schizophrenia spectrum disorders (also referred to here as “spectrum” disorders), as described in the 1968 Kety-led study, included “chronic schizophrenia” (which the researchers called “B1”), “acute schizophrenia reaction” (“B2”), “borderline schizophrenia” (“B3”; also called “latent schizophrenia”), “uncertain chronic schizophrenia” (“D1”), “uncertain acute schizophrenia reaction” (“D2”), “uncertain borderline schizophrenia” (“D3”), and “inadequate personality” (“C”; called “schizoid or inadequate personality” in the 1975 study). The researchers decided to count all these, including the “inadequate personality” and “uncertain” cases, as schizophrenia. The 1968 Danish-American schizophrenia spectrum is seen below.

AI. Definitely not schizophrenia

B1. Chronic Schizophrenia

B2. Acute Schizophrenia Reaction

B3. Borderline Schizophrenia

D1. Uncertain Chronic Schizophrenia

D2. Uncertain Acute Schizophrenia Reaction

D3. Uncertain Borderline Schizophrenia

C. Inadequate Personality

The 1968 Kety-led study index group consisted of 16 B1, 7 B2, and 10 B3 adoptees, for a total of 33. The 1975 study was based on the same index and control adoptees, but one index B3 diagnosis was changed to B1 on the basis of interview. The 1975 index group therefore consisted of 17 B1, 7 B2, and 9 B3 adoptees. The researchers grouped together the B diagnoses (B1, B2, and B3) and called them “definite schizophrenia.” They referred to the D diagnoses (D1, D2, and D3) collectively as “uncertain schizophrenia.” Figure 4 shows Kety and colleagues’ complete description of their Copenhagen spectrum categories, as seen in their original 1968 publication.

The “B” diagnostic categories captured a nebulous hodge-podge of behaviors, with the “D” diagnoses representing “uncertain” versions of these categories. In Kety and colleagues’ own words, “uncertain schizophrenia” was a “vague and subjective category, which hardly qualifies as schizophrenia according to our own or other criteria.”²⁸⁸ The “global” diagnostic system did not require adoptees or relatives to display a standard number of these behaviors, or even the same behaviors. The researchers used these descriptions to arrive at an overall diagnostic impression. The reliability and validity of these categories are highly questionable, and it could be argued that the Danish-American studies are invalidated on this basis alone.

In 1974 Kety claimed a “high degree” of reliability for the “definite” B diagnoses in his studies, although reliability was lower for the “uncertain” D diagnoses. He wrote that in contrast to claims by critics of psychiatry, who he said argued that “schizophrenia is a myth, that it has no biological substrate, and that it cannot be diagnosed reliably,” his results “indicate that it is possible to distinguish insanity quite readily from sanity, and, in fact, to distinguish a particular form of insanity called schizophrenia.”²⁸⁹ Kety had in mind critics within psychiatry such as Peter Breggin, R. D. Laing, and Thomas Szasz, as well the Rosenhan experiment, but he did not explain why it was necessary to perform an adoption study to determine whether psychiatrists are able to distinguish the “sane” from the “insane.”

Kety’s 1974 claim about the reliability of schizophrenia contrasted sharply with a conclusion he reached a year later. Here, Rosenthal, Kety, and colleagues recognized that psychiatric “studies of inter-rater reliability” up to that point “repeatedly turned up with relatively low figures, especially in the nonhospitalized populations.” Turning to their own Danish-American studies, although they claimed good agreement on determining the severity of a diagnosis, they recognized their inability to consistently agree on a diagnosis. “Taken together with previous reliability studies,” they wrote, “the data here indicate that independent judges may not agree regarding what [diagnosis] a person has, but they show remarkable agreement in deciding how severe ‘it’ is.”²⁹⁰

In Boyle’s view, Kety and colleagues’ spectrum descriptions were “arbitrary,” and the “criteria for inferring many of its concepts are extremely vague.” She noted that the researchers provided no data on the validity or reliability of these categories in 1968, nor did they indicate “how many of the characteristics on the list people must show in order to be called borderline schizophrenic or whatever.”²⁹¹ Psychiatrist Theodore Lidz commented in 1976 that the ability to distinguish between “a definite latent schizophrenic” and “an uncertain latent schizophrenic is a rather extraordinary feat” (italics in original).²⁹²

In the Kety-led 1968 and 1975 studies, the researchers selected as index adoptees only those with B “definite schizophrenia” conditions. Diagnoses among the relatives, on the other hand, included the B, D “uncertain schizophrenia,” and C “inadequate personality/schizoid” conditions. The researchers’ decision to expand the definition of schizophrenia in the four relative groups enabled them to find more spectrum diagnoses among these relatives. In their original 1968 and 1975 publications, Kety and colleagues implied that the issue was about fine-tuning the spectrum in the relative groups, when in fact it was about expanding it. In 1968 they wrote that when diagnosing relatives, “a system of classification with finer gradations than had been found useful in the selection of index cases would be needed.”²⁹³ In the 1975 study, they wrote that “psychiatric interviews on relatives outside of psychiatric institutions….permits a greater resolution into more specific diagnostic categories within what was designated as the ‘schizophrenia spectrum’ in the earlier [1968] study.”²⁹⁴ And in a later 1988 Provincial Study report, Kety and a co-author wrote, “For diagnoses in the relatives, a category of uncertain schizophrenia was necessary because a relative could not be rejected from the study as a candidate proband could be if the diagnosis were less than definite.”²⁹⁵ The researchers’ at times unclear writing and use of the passive voice, as seen in these passages and elsewhere, obscured the fact that they made important decisions at every step, and that the hidden decisions they made often had a major impact on their results. Although Kety and colleagues decided to define schizophrenia spectrum disorders much more broadly in the relative groups as compared with the adoptee (proband) group, they failed to state this clearly, and failed to provide sufficient justification for doing so.

In their final 1994 Provincial Study publication, Kety and colleagues dropped the “acute,” “uncertain,” and “inadequate personally” diagnoses from the schizophrenia spectrum (see below). The two remaining 1994 Provincial spectrum diagnoses were “chronic schizophrenia,” and the B3 equivalent “latent schizophrenia.” A total of 29 Provincial index adoptees were diagnosed with chronic schizophrenia, and 4 were diagnosed with latent schizophrenia. The final 1994 control group consisted of 24 adoptees and their biological and adoptive relatives.

Index Group Diagnostic Rates Must Be Significantly Higher than the General Population Rate

Kety argued that a control group constituted “the only legitimate basis for comparison” with the index group.²⁹⁶ However, if researchers find significantly more schizophrenia among index versus control relatives or adoptees—but at a rate not significantly higher than the general population expectation of roughly 1%—the results cannot be generalized to the non-adoptee population, nor can they be interpreted in favor of genetics. As Boyle put it, if “the index relatives resembled the general population and the control relatives were exceptionally free from diagnoses, then significant differences, but carrying very different interpretations, could appear.”²⁹⁷ Boyle concluded that “a simple comparison of two groups of biological relatives does not indicate how similar each is to the general population.”²⁹⁸

From a statistical perspective, if diagnoses among index adoptees or relatives are not significantly higher than the level found in the general population, diagnostic levels are the same in each population. In this case, the only valid conclusion one could draw from significantly higher index versus control diagnostic rates would be that the control adoptees or relatives experienced more schizophrenia-protective environments than experienced by their index counterparts. This finding would not support a conclusion that the higher index versus control schizophrenia rate is caused by genetic factors.

Rosenthal recognized that in order to “demonstrate that genes have anything to do with schizophrenia,” the “frequency of schizophrenia must be greater in the families of schizophrenics than in the families of nonschizophrenic controls or in the population at large.”²⁹⁹ The last part should have read “and in the population at large,” because a significantly higher index adoptee or relative schizophrenia rate versus controls cannot be interpreted in favor of genetics unless it is also significantly higher than the 1% general population rate.

The “Inadequate Personality” Diagnosis

The researchers based their Category C “inadequate personality” diagnosis on the 1968 DSM-II description. This diagnosis was not included in the 1980 DSM-III and subsequent editions. In the 1975 study, Category C was called “schizoid or inadequate personality.” The complete DSM-II description of inadequate personality reads as follows:

“This behavior pattern is characterized by ineffectual responses to emotional, social, intellectual and physical demands. While the patient seems neither physically nor mentally deficient, he does manifest inadaptability, ineptness, poor judgement, social instability, and lack of physical and emotional stamina.”³⁰⁰

In their 1968 and 1975 statistical calculations, Kety and colleagues counted relatives falling into this vague and non-psychotic category as “schizophrenic,” a practice that few mainstream authors have commented upon or criticized. The Rosenthal-led and Wender-led studies also counted “schizoid or inadequate personality” as a schizophrenia spectrum disorder.

In 1979, DSM-III architect Robert Spitzer and Jean Endicott examined the Danish-American records in an effort to define diagnostic categories for DSM-III.³⁰¹ They created a new diagnosis, “schizotypal personality disorder” (SPD), from eight symptoms distinguishing Kety and colleagues’ spectrum and non-spectrum relatives. (Kety subsequently described schizotypal personality disorder as “comparable to our diagnosis of latent schizophrenia.”³⁰²) Spitzer and Endicott, however, could not adequately distinguish the schizoid and schizotypal personalities as two discrete diagnostic categories. They concluded that SPD “was merely a subdivision of what has for years been referred to as Schizoid Personality Disorder.”³⁰³ The DSM-III differentiated schizoid personality disorder and schizotypal personality disorder on the sole basis of the latter’s “eccentricities of communication or behavior.”³⁰⁴ And in fact, in reference to the B3, D3, and C diagnoses, Kety, Rosenthal, and Wender wrote in 1978 that “it is doubtful that we could demonstrate a significant differentiation between these categories.”³⁰⁵ That is, they made the remarkable admission that they could not differentiate between a schizophrenia spectrum disorder (B3 and D3) and a non-spectrum disorder (C). And yet, the study depended on their ability to make such a distinction.

“Uncertain” Diagnoses in the Kety-Led Copenhagen Study

“Uncertain” diagnoses (D1, D2, and D3) should not have been counted as “schizophrenia” in these studies, and certainly not the triply vague “uncertain borderline schizophrenia” (D3) diagnosis. If the schizophrenia concept during this era was, as Szasz put it, “wonderfully vague in its content,” then “borderline schizophrenia” was a questionable version of a vague diagnosis, and “uncertain borderline schizophrenia” was an uncertain version of a questionable version of a vague diagnosis.

According to Kety, “In the case of the relatives, questionable or uncertain schizophrenia had to be added if relatives with less certain diagnoses were not to be lost.”³⁰⁶ Perhaps so, but Kety could have prevented these relatives from becoming “lost” without counting them as schizophrenia in his study’s statistical calculations. Genetically biased researchers engaging in p-hacking can use their “degrees of freedom” to help ensure that they reach the desired conclusions. If a diagnosis helps support genetic positions, they can decide to count it; if it doesn’t, they can decide not to count it.

In the 1975 Kety-led study, “uncertain” diagnoses represented 35% (13/37) of all index biological relative spectrum diagnoses, and another 35% were diagnosed with “schizoid or inadequate personality” (Category C). Together, these “uncertain” and “inadequate personality” diagnoses accounted for a whopping 70% (26/37) of all index biological relative spectrum diagnoses.³⁰⁷

Homophobic and Non-Psychotic Diagnostic Criteria

For Rosenthal, it was enough for a person to be “cold, distant and inadequate, or odd and eccentric” to qualify as having a “schizophrenic-like type of disorder.”³⁰⁸ Being judged as a “pervert,”³⁰⁹ or as being “perverse (homosexual, transvestite)” also qualified.³¹⁰ Kety, Rosenthal, and Wender referred to same-sex sexual behavior or orientation as a symptom of schizophrenia in all three of their original 1968 “definite schizophrenia” diagnoses (see Figure 4). The B1 diagnosis was characterized by “few heterosexual contacts,” B2 included “homosexual panic,” and B3 included a “mixture of hereto- and homosexuality.”³¹¹ Although Kety and colleagues based these diagnoses on 1968 DSM-II diagnostic descriptions, the DSM-II did not mention anything about sexual orientation or behavior in its descriptions of the various types of schizophrenia, nor has any other version of the DSM.³¹²

For Wender, “schizophrenia-related illnesses” contained people who “vary from the shy, timid and unsociable to the callous, cold, harsh, and distant, from the quiet, empty, and intelligent to the sensitive and poetic or to the militant, rigid, and fanatic (political or religious).”³¹³ One can only wonder how many “shy, timid and unsociable,” “odd and eccentric,” “sensitive and poetic,” “perverted,” “homosexual,” or “rigid” people Rosenthal, Kety, and Wender diagnosed with a “schizophrenia spectrum disorder” (B3, D3, or C) in the Danish-American studies.

Decades later, a 2016 team of molecular genetic researchers typically characterized schizophrenia as being “defined by psychotic experiences (hallucinations, delusions, and thought disorder) and negative symptoms (e.g., blunted affect and apathy).”³¹⁴ These researchers were apparently unaware of the irony that the earlier adoption researchers, upon whose work schizophrenia gene searches are in part based, defined “schizophrenia” very differently.

No Significant Elevation of B1 Chronic Schizophrenia in 1968 or 1971

The 1968 Kety-led study found zero cases of chronic schizophrenia (B1) among the 65 identified first-degree biological relatives of the index adoptees, and the 1971 Rosenthal-led study found that only 1 of the 76 adopted-away biological offspring of a parent diagnosed with a schizophrenia spectrum disorder had received a hospital diagnosis of chronic schizophrenia. Had the researchers decided to count only these chronic cases—as schizophrenia was defined in Denmark—they would have found no evidence that schizophrenia has a genetic component.³¹⁵ As Rosenthal subsequently admitted, “If we had relied only on hard-core, process [chronic schizophrenia] cases, we would have found no significant difference between our index and control subjects.”³¹⁶

Interviews and “Pseudointerviews” in the Kety-Led 1975 Study

Substandard interviews were used to make diagnoses in the 1975 Kety-led study. Some of these “interviews” never took place, but instead were fabricated by the investigators on the basis of hospital records. In the raw data Kety called these “pseudointerviews,” but no mention of this practice appeared in any Danish-American publication. According to Kendler and Gruenberg, the Danish-American researchers constructed “detailed pseudointerviews…for all of the index adoptees.”³¹⁷ Lewontin, Rose, and Kamin, the authors of Not in Our Genes, had been in correspondence with one of the psychiatrists conducting the interviews. According to Lewontin and colleagues,

“In several cases, when relatives were dead or unavailable, the psychiatrist ‘prepared a so-called pseudo interview from the existing hospital records.’ That is, the psychiatrist filled out the interview form in the way in which he guessed the relative would have answered” (italics added).³¹⁸

The researchers apparently used these pseudointerviews when diagnosing index adoptees and unavailable relatives, and it borders on scientific scandal that Kety and colleagues failed to mention anything about these non-existent “interviews” in their numerous publications. Psychologist David Jacobs called this practice “outright fraud.”³¹⁹

Of the interviews the researchers did conduct, they believed that they could make a psychiatric evaluation of a reluctant interviewee on the basis of a five-minute doorstep conversation.³²⁰

Chronic B1 Schizophrenia in the 1975 Kety-Led Study

In the 1975 Kety-led study, the researchers reported a statistically significant B1 chronic schizophrenia rate among their index biological relatives versus their control biological relatives. They counted five B1 diagnoses among these 173 index biological relatives, versus zero B1 diagnoses among the 174 control biological relatives. The index B1 rate was reported as statistically significant at the .03 level. However, four of the five index diagnoses were given to half-siblings (two maternal, two paternal), which runs counter to genetic predictions.³²¹ In addition, this 5/173 index rate is not significantly higher than the population expectation of at least 1/173.

It turns out that, even counting the half-siblings, the five index B1 biological relative diagnoses did not constitute a statistically significant clustering versus controls. Kety and colleagues diagnosed the biological father of control adoptee “C9” as B1 in 1968 on the basis of records, but he had died before he could be interviewed for the 1975 study. They arbitrarily decided to change his 1968 B1 diagnosis to no diagnosis in 1975, possibly on the basis of his “pseudointerview,” but they provided no justification for removing his B1 diagnosis. ³²² Although Kety and colleagues’ 1975 calculations made it appear as though B1 diagnoses were significantly concentrated among the index biological relatives versus the control biological relatives (5 to 0), the difference would have been statistically non-significant (5 to 1) if they had decided to count this 1968 B1-diagnosed control relative.³²³

The Necessity of “Broadening the Concept of Schizophrenic Disorder as Widely as it May Have Ever Been Reasonably Conceived Before”

While citing other reasons in their most important publications, it is clear that the main reason that the researchers decided to broaden the definition of schizophrenia to include “schizophrenia spectrum disorders” was to have enough cases to be able to conduct their studies. In a little-known 1971 article, Rosenthal basically admitted as much:

“The second [important] feature [of the research] has to do with the fact that we have included a broad spectrum of disorders in the ones I am calling schizophrenic. These include not only the classical chronic, process types of cases, but patients called doubtful schizophrenic, reactive, schizo-affective, borderline or pseudoneurotic schizophrenic, or schizoid or paranoid. If we dealt only with hardcore [B1, chronic] schizophrenia, our ns [number of participants/subjects] would have been too small to make any of these studies meaningful” (italics added).³²⁴

Rosenthal also acknowledged elsewhere, “It seems somewhat ironic that…Paul Wender and I…in concert with Seymour Kety [were] in effect broadening the concept of schizophrenic disorder as widely as it may have ever been reasonably conceived before.” The researchers, according to Rosenthal’s account, “strained to encompass all disorders that shared salient clinical and behavioral manifestations with process [chronic B1] schizophrenia and to group these as a spectrum of schizophrenic disorder.”³²⁵ Yet in another publication Rosenthal admitted that, prior to the Danish-American studies, “In almost all genetic studies of schizophrenia, the investigators based their evaluations of the mode of inheritance on cases diagnosed as clear-cut [chronic] schizophrenia.”³²⁶ Indeed, between 1900 and 1960, symptoms such as “thought disorder,” “delusions,” and “hallucinations” were included in all diagnostic systems of schizophrenia and dementia praecox.³²⁷

The necessity of creating a spectrum was due to the investigators finding, in the 1968 Kety-led study, only 16 cases (about 0.3%) of B1 chronic schizophrenia among the 5,483 identified adoptees, when 38-40 cases (about 0.7%) would have been expected on the basis of population statistics.³²⁸ This sample of 16 chronic schizophrenia adoptees was, as Rosenthal acknowledged, a “lower than expected yield,”³²⁹ which was “too small to make any of these studies meaningful.”³³⁰ Even this “lower than expected yield” of 16 index B1 adoptees may have been too high. In their 1984 “independent analysis” of the Kety-led Copenhagen Study based on “operational” DSM-III diagnostic criteria, Kendler and Gruenberg diagnosed only 11 of these 16 adoptees (69%) with DSM-III schizophrenia.³³¹

This unexpectedly low adoptee schizophrenia rate suggested that simply being reared by adoptive parents, who had been screened by Danish adoption agencies and courts for mental health and economic stability, had reduced the schizophrenia rate by over 60%.³³² As critics Theodore Sarbin and James Mancuso concluded, the Danish “case workers had apparently done an excellent job in screening adoptive families and had placed these children with unusually fine families.”³³³ Thus, in the beginning phases of their work the researchers uncovered—yet overlooked—important evidence suggesting that being reared in a stable family environment greatly reduced one’s chances of being diagnosed with schizophrenia later in life.

The researchers first described their spectrum in the 1968 Kety-led publication, and subsequently claimed that they had developed the spectrum hypothesis at an earlier stage of the research.³³⁴ However, because there is no published evidence appearing prior to 1968 supporting the researchers’ claim that they decided early-on to broaden the definition of schizophrenia “as widely as it may have ever been reasonably conceived before,” and because their later accounts are contradictory, the evidence suggests that the researchers “hypothesized after the results were known” and developed the spectrum hypothesis after finding an unexpectedly low number of B1 diagnosed adoptees and relatives.

“Necessarily Vague” Diagnostic Boundaries

Schizophrenia and the spectrum disorders were poorly defined by the investigators, with Kety admitting, amazingly, that their diagnostic descriptions were “necessarily vague,”³³⁵ with a “lack of sharp boundaries.”³³⁶ Kety and colleagues frequently claimed that they based their diagnoses on Eugen Bleuler’s original 1911 descriptions of schizophrenia and “latent schizophrenia,” and they used the terms “borderline schizophrenia” and “latent schizophrenia” interchangeably in the Copenhagen Study. Bleuler, however, believed that “only a few isolated psychotic symptoms can be utilized in recognizing” schizophrenia, and therefore demanded a “very high diagnostic threshold value” for identifying it.³³⁷

For Bleuler, mild or “simple” schizophrenia (a B3 equivalent, see Figure 4) was a retrospective diagnosis to be made only on the basis of a person’s later diagnosis of schizophrenia:

“Such mild cases are often considered to be ‘nervous’ or ‘degenerated’ individuals, etc. But if we follow the anamnesis of those who are admitted to the hospital in later years because of an exacerbation of their difficulties, a criminal charge, a pathological drinking bout or some such episode, we can usually find throughout the entire past history of the individual mildly pathological symptoms which in the light of their recent illness unquestionably have to be considered as schizophrenic” (italics added).³³⁸

As this passage shows, Bleuler recognized B3-type behavior as relevant only “in the light of” a person’s “recent” diagnosis of “schizophrenic” illness. And according to Kety, Bleuler saw “latent schizophrenia” as an even “broader and milder” form of “simple schizophrenia.”³³⁹

The Arbitrary Inclusion of “Borderline Schizophrenia” in the Spectrum

The investigators claimed that the Kety-led 1968 and 1975 results showed that “borderline schizophrenia” (B3) was genetically related to chronic schizophrenia, because it was significantly concentrated in the index biological relative group compared with the control group. However, they used improper methods of counting and combining diagnoses to arrive at this conclusion. As I showed in Chapter 3 of The Missing Gene, an examination of their results, based on proper methods of counting, leads to the opposite conclusion.³⁴⁰

In the 1975 Kety-led study, there was no statistically significant elevation of B3 “borderline schizophrenia” in the index versus control biological relative groups.³⁴¹ Kety and colleagues found statistically significant results only by combining this diagnosis with chronic schizophrenia and the “uncertain” spectrum diagnoses, and then concluding that all were genetically related to chronic schizophrenia. In doing so, they skipped a crucial step in determining the relationship between chronic and borderline schizophrenia. That is, they decided that individual diagnoses, standing alone, did not have to cluster significantly among their index versus control biological relatives.

From the genetic perspective, Kety, Rosenthal, Wender, and colleagues had a vital interest in retaining B3 in the spectrum. Had they decided to remove this diagnosis, their 1968 and 1975 index groups would have been reduced to the original 23-24 B1 and B2 adoptees, or possibly only the 16-17 B1 adoptees. A decision to remove B3 (and therefore D3) would have either meant the unsuccessful conclusion of their work, or the need to identify thousands more Danish adoptees in the hope of finding enough B1 cases to be able to continue the study. It is noteworthy that in their 1984 analysis, Kendler and Gruenberg diagnosed only one of the original ten B3 adoptees (10%) with schizotypal personality disorder, the DSM-III equivalent of B3.³⁴²

In a 1988 report on the still-in-progress Provincial Study results, Kety and Loring Ingraham wrote:

“At the prototypical end of the spectrum, chronic schizophrenia is found exclusively in the biological relatives of chronic schizophrenia patients where it occurs at a low prevalence (approximately 3%), whereas the prevalence in the biological relatives in the normal controls is negligible. The same is true for uncertain chronic schizophrenia. Latent or borderline schizophrenia was found at a 4-5% prevalence in the biological index relatives and 1% to 1.5% in the biological relatives of controls. This is also true where the symptoms are less distinct and the diagnosis is designated uncertain. Since neither in chronic nor in latent schizophrenia the results for the definite or uncertain diagnoses are statistically different, it appears justified to combine them” (italics added).³⁴³

According to Kety, in order to achieve statistically significant findings it appeared “justified” to combine spectrum disorders in his study’s statistical comparisons. However, he could just as easily have decided to conclude that, “since neither in chronic nor in latent schizophrenia the results for the definite or uncertain diagnoses are statistically different,” the Provincial Study found no evidence that chronic schizophrenia, or latent schizophrenia, had a genetic component.

The Arbitrary Exclusion of “Acute Schizophrenia” from the Spectrum

Kety, Rosenthal, and colleagues, on the other hand, concluded that “acute schizophrenia” (B2) was not a “subtype of schizophrenia,” and decided to exclude this diagnosis from the spectrum. From a statistical perspective, however, there was equal justification in their results for excluding “borderline schizophrenia” (B3) as well, but they arbitrarily chose not to do so.³⁴⁴ Kety cited E. Bleuler in support of removing B2 from the spectrum, arguing that the DSM-II had “deviated” from Bleuler’s teachings by including “‘acute schizophrenic reaction’…despite Bleuler’s admonition that these are ‘partial phenomena of the most varied diseases [whose] presence is often helpful in making the diagnosis of a psychosis, but not in diagnosing the presence of schizophrenia.’”³⁴⁵ Kety thereby cited Bleuler’s high diagnostic threshold in support of the researchers’ decision to remove B2 from the spectrum, while ignoring a comparable threshold Bleuler had established for B3.

High Rate of Spectrum Diagnoses Among Control Biological Relatives

The 1975 Kety-led study reported an 11% schizophrenia spectrum disorder rate (19/174) among all control biological relatives.³⁴⁶ Based on the prevailing Danish chronic (B1) schizophrenia population expectation of about 0.7%, by chance we would have expected only one or two of these 174 biological relatives of non-diagnosed control adoptees to have been diagnosed with chronic schizophrenia.³⁴⁷ In the Rosenthal-led study, the researchers diagnosed 26% (16/62) of the control adoptees with a consensus spectrum disorder.³⁴⁸ This indicates that the Danish-American researchers expanded the definition of schizophrenia so widely, that it produced a 10- to 15-fold increase of “schizophrenia” in Denmark.

Attachment Rupture: The Use of Late-Separated and Late-Placed Adoptees

The researchers decided to include many late-separated and late-placed adoptees (abandoned children) in their samples. In order to reduce attachment disruption trauma and the post-natal influences of the biological parents, a psychiatric adoption study should, at the very least, be limited to children who are separated from their birthmother at or very soon after birth. The transfer to their adoptive homes should occur very soon thereafter. As Faraone and his psychiatric geneticist colleagues noted, “If the child has lived with a parent for even a short period of time prior to adoption, the biological relationship will have been ‘contaminated’ by environmental factors.” They recognized that “some might even argue that the child’s contact with the mother immediately after birth creates a residue of environmental influence” that could lead to “subsequent psychopathology.”³⁴⁹

Of the 33 index adoptees (probands) in the Kety-led 1968 and 1975 studies, 19 were separated from their biological parents within one month, 6 from one to three months, 2 from three months to six months, 5 from six months to twelve months, and one remained with the parent(s) for “more than two years.” The average amount of time that an index adoptee spent in a Danish “children’s institution” (orphanage) was 10.2 months, and he or she spent an average of 3.6 months in a foster home. The average age at transfer to the adoptive parents was 18.3 months. The average age at legal adoption was 38 months.³⁵⁰ In the Rosenthal-led study, the “transfer-age” ranged from 5 days to 48 months. The median transfer age was 5.9 months.³⁵¹

In the 1968 Kety-led study, the investigators wrote that “one troublesome problem” in their study

“was the presence among the 33 index cases of a number who had lived with their biological families for various periods from one month to one year or even more. Although these made the hoped-for separation of hereditary from environmental factors less rigorous, we had decided not to exclude such cases because of the small size of our index group.”³⁵²

One can only imagine the attachment rupture trauma experienced by these children, the potential impact of which cannot be dismissed simply because the researchers’ sample was not large enough to achieve the “hoped-for separation” of nature and nurture, or because they used a matched control group.

In contrast, in the Colorado Adoption Project (CAP), a longitudinal American behavioral genetic study of psychological and behavioral characteristics initiated in the 1970s, all adoptees were separated from their biological mothers at or shortly after birth, and were placed into qualified adoptive homes within a few weeks.³⁵³ Even so, future adoptees spent nine months in the prenatal environment of their birthmother, meaning that the separation of birthparent and adoptive family environments can never be complete. (Like other adoption studies, the CAP has its own set of problem areas.³⁵⁴)

In a rare psychiatric adoption study description of the conditions that the children under study endured in early-to-mid 20^th century orphanages (“children’s institutions,” “foundling homes”), Heston recognized that poor conditions, and the lack of adult nurture, had a negative impact on children’s psychological development:

“The actual extent to which these children were deprived of maternal or emotional nurture beyond that implicit in group care must be inferred largely from indirect evidence… . The later history of these children strongly supports the hypothesis that significant deprivation did occur. Random observations recorded in school or nursery records and recollections of foster parents and subjects describe several children as shy, withdrawn, demanding excessive attention, or sad; few as happy, spontaneous, or normal. Some were requiring sedation at night… . It is certain that most of these children were unhappy and probable that there was significant deprivation of emotional nurture.”³⁵⁵

As Heston described it, and as most people would expect, children abandoned and placed in an orphanage suffered greatly. Many were unhappy, many experienced a “significant deprivation of emotional nurture,” and some had to be sedated at night.

During sensitive developmental periods, therefore, in addition to spending nine months in their biological mother’s prenatal environment, most children studied in schizophrenia adoption research (1) were reared for a certain period of time by their biological parent(s), (2) suffered a disruption of attachment bonds with their biological parent(s), and/or (3) were placed into unstable or psychologically/developmentally harmful environments, such as foster homes and orphanages, between separation and adoption.

Violation of the Statistical Assumption of Independent Observations

In the Kety-led studies, critic Lorna Benjamin noted in 1976 that the “procedure of counting up all the possible relatives of each index case and pooling them as if they were independent samples … would allow some families to disproportionately affect the results.”³⁵⁶ Although Kety and colleagues did address this issue, their decision to emphasize the diagnostic rate among individual relatives, as opposed to individual families, violated the assumption of independent observations underlying the statistical methods they used. Because family environments may play a role in causing schizophrenia and psychosis, genetic study diagnoses given to people raised in the same family environment are not independent observations, because they shared this environmental factor in common.

In an unpublished 1975 letter, Loren Mosher, then editor of Schizophrenia Bulletin and the Director of Schizophrenia Research at the U.S. National Institute of Mental Health, weighed in on this issue:

“The actual sample size of all the adoption studies is, at least for genetic purposes, the number of index probands [adoptees in this case], not the number of relatives identified. The power of the adoption methodology is its separation of heredity and environment for genetic analysis; therefore, when N’s are reported as the number of biological relatives seen, it leaves the misleading impression that the genetic/environmental separation is applicable in this group, whereas, in point of fact, it is not. Basically, studying either biological or adoptive relatives is just a special family study” (italics in original).³⁵⁷

Kety and colleagues’ mistaken practice of counting relatives who grew up in the same family as independent observations gave the “misleading impression” that they had successfully separated genetic and environmental influences. For Mosher, this practice transformed the investigation into little more than a “special” type of family study.

In their 1975 publication, Kety and colleagues presented a table showing that more index biological families contained at least one spectrum diagnosis versus the control biological families, at a statistically significant level. The table listed 14 index families versus 3 control families with at least one B diagnosis, and 17 families versus 5 families with at least one B or D diagnosis.³⁵⁸ As Boyle pointed out in 2002, however, Kety and colleagues miscounted the number of index group biological families with at least one B diagnosis as 14. The correct number is actually 8 affected families.³⁵⁹ When Kety’s diagnostic comparisons are adjusted to correct for this error, neither of the above comparisons is statistically significant.³⁶⁰ For Boyle, it “is very difficult to understand how this mistake…was made in the first place.” She noted that the error has “not been corrected,” and that the incorrect figures continue to be cited.³⁶¹ Indeed, Kety, Rosenthal and colleagues continued to cite these erroneous family diagnostic rates in a 1978 publication.³⁶²

Counting “Manic-Depression” as a Schizophrenia Spectrum Disorder

In the Rosenthal-led studies (but not in the Kety-led studies), the researchers counted “manic-depression” as a schizophrenia spectrum disorder³⁶³ despite Rosenthal’s insistence elsewhere that schizophrenia and manic-depression are “genetically distinct and different disorders.”³⁶⁴ Without these manic-depressive cases, Rosenthal and colleagues would not have been able to claim statistically significant results in the genetic direction in their 1971 publication.³⁶⁵

Even though Rosenthal viewed schizophrenia and manic-depression (now known as bipolar disorder) as genetically distinct and different disorders, the 1971 Rosenthal-led study listed “manic-depressive psychosis” as one of the diagnoses “that we are tentatively including in the ‘schizophrenia spectrum.’”³⁶⁶ This did not prevent Rosenthal, Kety and colleagues from responding to critics five years later with the false claim that “manic-depressive illness was never thought to be in the schizophrenia spectrum by us,” and that the Rosenthal-led study had included manic-depressive cases only as a “comparison pathology group.”³⁶⁷

The researchers diagnosed 7 of the 76 (9%) index biological parents with manic-depressive disorder, and one or more of them diagnosed an additional 17 parents (22%) with this condition.³⁶⁸ Why did this “genetically distinct and different disorder” qualify a parent as a schizophrenia index case? Rosenthal offered two reasons in 1968:

“These [manic-depressive] cases were included for two reasons. The first was one of expediency. There were periods when we simply did not have enough schizophrenic parents processed and the staff in Copenhagen had no subjects to examine. The second and more important reason derived from this question: What if we should find differences between our Index and Control groups… . If we had a comparison pathology group, we might be able to learn something… . about the possible genetic relationship between schizophrenia and manic-depressive psychosis” (italics added).³⁶⁹

Rosenthal and colleagues thereby included an admittedly non-related diagnosis in a schizophrenia genetic study for reasons of “expediency.” Imagine the folly of a study on the genetics of heart disease that added liver disease patients because the staff “had no heart disease subjects to examine,” and then decided to count liver disease as a “heart disease spectrum disorder.” In the words of psychologist Alvin Pam, a critic of this study, the inclusion of manic-depressive parents rendered the study “invalid on its face.”³⁷⁰ Lidz and his colleagues argued that “the study could no longer properly be termed a study of adopted-away offspring of schizophrenic parents.”³⁷¹ Indeed, it couldn’t.

The Kety-Led Studies: Changing the Research Design After the Results Came In

In this section I will provide evidence that the researchers in the Kety-led study abandoned a planned group comparison that did not produce statistically significant results, in favor of a different group comparison that produced such results. My purpose here is to discuss the comparisons the researchers decided to make, rather than to speculate about which comparisons they should have made.

The 1968 Kety-Led Study. We have seen that the Kety-led studies began in 1963, and that the records-based results were first published in 1968. Although in 1968 Kety and colleagues concluded in favor of genetics on the basis of finding a significantly higher rate of spectrum diagnoses in the index biological (IB) versus the control biological (CB) relative groups, they appear to have changed the study’s design after the first relative group comparisons did not produce statistically significant results. The four relative groups are seen in Figure 5 (which is a simplified version of the relative groups shown in Figure 2).

In his 1959 Science article, Kety envisioned an adoption study that would compare diagnostic rates between the biological and adoptive relatives of people diagnosed with schizophrenia:

“Another possible means of better controlling the environmental variables would be to make a careful study of schizophrenia in adopted children, with comparison of the incidence in blood relatives [IB] and in foster relatives [IA]. Perhaps only a survey on a national scale would provide the requisite numbers of cases for any of these studies.”³⁷²

In a 1967 publication, which was based on a paper he presented at the March 29-31, 1967 First Rochester International Conference on Schizophrenia, Rosenthal described a comparison similar to Kety’s 1959 description:

“In Denmark, with the collaboration of Dr. Fini Schulsinger and others, we began with adoptees who are now schizophrenic. We compare the incidence of schizophrenic disorders in their biological [IB] and adoptive [IA] families. The same procedure is carried out for a matched group of normal adoptees [CB versus CA], who serve as

controls.”³⁷³

Rosenthal is therefore on record as stating, as late as March, 1967, that he and his colleagues intended to compare groups IB versus IA (index adoptive) relatives or families, and then to compare groups CB versus CA (control adoptive) relatives or families (see Figure 5). However, they did not begin their analyses of the data until April of 1967.³⁷⁴ Kety and colleagues presented the results to their psychiatric genetics colleagues (and others) two months later at the Dorado, Puerto Rico conference, to be published the following year in The Transmission of Schizophrenia.

Although Kety and Rosenthal were on record prior to 1968 as intending to base the study on IB-IA comparisons—and I am aware of no other published descriptions of their methods and group comparisons appearing prior to 1968—in their 1968 publication they explained why they decided against making this comparison:

“The biological [IB] and adoptive [IA] parents differ in age, socio-economic class and in the particular selective processes inherent in their having become biological or adoptive parents, making comparisons difficult between them with respect to the prevalence of mental illness.”³⁷⁵

They wrote in the 1968 study that because IB-IA comparisons were “difficult,” they decided to base their conclusions on IB-CB comparisons:

“The prevalence of particular types of mental illness in each group of relatives of the index cases can appropriately be compared with that in the corresponding relatives of the controls [IB versus CB, IA versus CA], permitting the separate testing of hypotheses based on genetic or environmental factors in the transmission of schizophrenia.”³⁷⁶

The 1968 Kety-led study reported 13 schizophrenia spectrum disorders out of 150 IB relatives (8.7%), and 2 such disorders out of 74 index adoptive (IA) relatives (2.7%).³⁷⁷ As seen in Figure 6, this IB versus IA comparison is not statistically significant, meaning that if the researchers had decided to stick with the comparison described by Kety in 1959, and by Rosenthal in 1967, they would have had to conclude that their study found no evidence in favor of genetic influences on schizophrenia. The decision to abandon the non-significant IB-IA comparison in favor of the significant IB-CB comparison transformed the study from one that found no genetic influences on schizophrenia spectrum disorders, into one that supposedly found such influences. This is another example of “hypothesizing after the results are known” (HARKing), because the evidence suggests that the researchers’ original intention had been to compare groups IB and IA to determine whether genetic factors influence or cause schizophrenia. After the results were known to them, they switched to the statistically significant IB-CB comparison, and implied that they had planned this comparison all along.

Although the diagnostic percentages seen in Figure 6 showed a higher IA rate compared to the CB rate (2.7% versus 1.9%, respectively), the lower number of IA relatives (74) led to an IB-IA comparison that did not reach statistical significance, whereas the IB-CB comparison did reach statistical significance due to the higher number of CB relatives (156).

Critics Theodore Lidz and Sidney Blatt wrote in the April, 1983 edition of the American Journal of Psychiatry, correctly as we have seen, that the “original purpose” of the Kety-led study had been “to differentiate genetic from intrafamilial environmental factors by comparing the occurrence of [“schizophrenic”] disorders in the biological [IB] and adoptive [IA] relatives of schizophrenic patients who had been adopted at a very early age.”³⁷⁸ Theodore Lidz (1910-2001) was a psychiatrist and Yale University professor interested in the role of the family environment in causing schizophrenia, and was the lead author of the 1965 book Schizophrenia and the Family. Sidney Blatt (1928-2014) was a Yale psychologist interested in personality development. Lidz attended the March, 1967 Rochester schizophrenia conference, and may well have listened to Rosenthal’s talk describing the Kety-led study plan to test genetic theories on the basis of the IB-IA comparison.³⁷⁹ I will examine Kety’s response to Lidz and Blatt’s assertion shortly.

Without acknowledging that his 1967 description of the researchers’ intended comparison groups had changed, in his 1970 book Genetic Theory and Abnormal Behavior Rosenthal described the revised comparison, which became the basis of the Kety-led studies:

“A higher incidence among the biological relatives of index cases [IB] than of controls [CB] indicates that heredity is contributing significantly to the disorder.”³⁸⁰

It could be argued that Kety and Rosenthal each had differing ideas about which groups to compare in the Kety-led studies, and that Kety’s ideas won out before the results were analyzed. This is unlikely, and like most other areas of behavioral research the investigators were able to present their methods, results, and conclusions for the first time in the same 1968 publication, without any prior public record of their intended methods and procedures other than the passages I have quoted here. This is a widespread problem, certainly not unique to schizophrenia adoption research, which we have seen would be greatly reduced by requiring studies to be preregistered.

The 1975 Kety-Led Study. Kety and colleagues, therefore, based their 1968 records-based conclusions on the statistically significant IB-CB comparison, while explaining that they did not make the statistically non-significant IB-IA comparison because making this comparison was “difficult.” However, when the interview-based study’s IB-IA comparison later became statistically significant in the genetic direction due to a larger sample, and additional diagnoses based on interviews, in a 1974 American Journal of Psychiatry article Kety appeared to cite the now-significant IB-IA comparison in favor of genetics:

“For any of these diagnoses of schizophrenic illness, the prevalence in those genetically related to the schizophrenic index cases [IB] is 13.9 percent compared to 2.7 percent in their adoptive relatives [IA] or 3.8 percent in all subjects not genetically related to an index case [IA, CB, and CA] (see table 2). These differences between the group genetically related to the schizophrenic index cases and those not so related are highly significant statistically and speak for the operation of genetic factors in the transmission of schizophrenia” (italics added).³⁸¹

In this 1974 passage, Kety compared the 13.9% IB rate to the 2.7% IA rate, and then compared the 13.9% IB rate to the 3.8% rate among “all subjects not genetically related to an index case” (IA, CB, plus CA). He then stated that the differences between these rates were “highly significant statistically.” In using the word “or,” he appeared to distinguish the IB versus IA comparison from the IA versus the combined IA, CB, and CA comparison, and then claimed that in both comparisons the group difference was statistically significant in the genetic direction. In a 1974 conference presentation, Kety again stated that the IB spectrum rate “is 13.9% compared to 2.7% in their adoptive relatives or 3.8% in all subjects not genetically related to an index case.”³⁸²

The 1975 Kety-led interview-based publication also worked the IA relative rate into the argument in favor of genetics:

“If we consider the total schizophrenia spectrum…we find that 21 percent of the biological index relatives [IB] fell within that category, as compared to 11 percent of the biological relatives of the controls [CB], and 5 and 8 percent respectively of the adoptive relatives [IA, CA]. There is thus a highly significant concentration of schizophrenia spectrum disorders in the biological relatives of index cases” (italics added).³⁸³

In 1976, however, one year into the Provincial Study, Kety and colleagues wrote that “another type of inappropriate comparison that some have made is that between adoptive and biological relatives,” implying that they had never planned or made such a comparison.³⁸⁴

1978 Chapters. In a 1978 chapter in the edited book The Nature of Schizophrenia, Kety and colleagues based their conclusions on IB-CB comparisons.³⁸⁵ In another 1978 chapter, they made the IB-CB comparison, and also compared IB diagnoses to the other three groups combined (IA, CB, and CA).³⁸⁶ In Kety’s solo-authored 1978 chapter in The Harvard Guide to Modern Psychiatry, he alluded to IB-IA comparisons and highlighted spectrum diagnostic rates between these groups (13.9% IB, 2.7% IA), and claimed a statistically significant difference between the IB relatives versus the combined IA, CB, and CA relatives.³⁸⁷ In another 1978 Kety publication, he made similar comparisons.³⁸⁸

Kety’s 1983 Position. In the June, 1983 edition of the American Journal of Psychiatry, Kety responded to Lidz and Blatt’s 1983 statement that his original intention had been to compare diagnoses between IB and IA relatives. Kety claimed that this objective “was not ours,” and wrote that “Lidz and Blatt have misunderstood ‘the original purpose of the project’ and the logic of our design.” Kety again came out against the practice of comparing IB versus IA diagnostic rates. He denied that he and his colleagues had planned to make that comparison, even though the Kety 1959 and Rosenthal 1967 publications clearly showed that they had. According to Kety,

“We anticipated that there would be differences between adoptive [IA] and biological [IB] relatives in age, socioeconomic status, life style, and other variables. For that reason we planned not to make comparisons between these two groups of relatives but, instead, as described fully in the original publications and outlined above, to compare each group with their respective controls [IB versus CB, IA versus CA] in evaluating separately the significance of genetic or family-related environmental factors.”³⁸⁹

In a second 1983 response to Lidz and Blatt, in the form of a letter to the editor of the American Journal of Psychiatry, Kety again denied that he and his colleagues had ever made diagnostic comparisons between their IB and IA groups. Referring to the 1975 interview-based study, he wrote,

“The prevalence of chronic, latent, and uncertain schizophrenia was much greater in the biological relatives of schizophrenic adoptees than in their adoptive relatives (13.9% versus 2.7%, respectively) and significantly so (p = .01, Fisher’s exact probability, two-tailed). We did not make that comparison, however, or compute its significance before this, because, for the reasons previously indicated (June 1983 issue) that comparison would have been improper and conclusions drawn from it fallacious.”³⁹⁰

Although we saw that Kety did appear to “make that comparison” in 1974 in American psychiatry’s leading journal, in this 1983 publication he wrote that IB-IA diagnostic comparisons were “improper,” and that conclusions drawn from such comparisons would be “fallacious.”

It appears that for Kety, with the results in hand, the planned IB versus IA comparison became “difficult,” “inappropriate,” “improper,” and “fallacious” when it failed to support the genetic position in 1968, but spoke “for the operation of genetic factors” in 1974 when it appeared to him to support this position. A few years later, with the Provincial Study in full swing, he denied that he and his colleagues had ever planned or made IB-IA comparisons.

Lidz and Blatt were among the few insiders to point to the glaring errors, biases, and contradictions in the Danish-American studies, but their views were marginalized by mainstream psychiatry at a time when the field was attempting to establish itself on medical model principles while in the midst of a rhetorical “biological revolution in psychiatry.” The medical model was celebrated in Andreasen’s 1984 book The Broken Brain, where she asserted that the major psychiatric disorders “are…diseases caused principally by biological factors, and most of these factors reside in the brain.”³⁹¹

***

We have seen that in addition to manipulating the definition of schizophrenia in support of genetic positions, the evidence suggests that the researchers changed their hypotheses and planned comparison groups in the 1968 Kety-led study, after the results were known to them (HARKing). This was a genetically biased violation of a “cardinal rule in experimental design,” which states “that any decision regarding the treatment of data must be made prior to an inspection of the data.”³⁹²

The Rosenthal-Led Study: Assuming the Genetic Basis of Schizophrenia in the Process of Testing for It

Although the purpose of the Danish-American adoption studies was to determine whether genetic factors play a role in causing schizophrenia, the spectrum concept itself implicitly assumed that chronic B1 schizophrenia is genetically based. The researchers then attempted to determine whether the various spectrum disorders were “genetically related” to an already assumed-to-be-genetic “chronic schizophrenia.”

The Rosenthal-led study was actually two studies. As Lidz, Blatt, and Barry Cook pointed out in their 1981 critique, it consisted of a 1968 sample, plus a 1971 sample. Rosenthal and colleagues claimed no statistically significant findings on the basis of their 1968 sample, while stressing that their results were “preliminary” and that “the N is small and will increase considerably.” ³⁹³ “The figures presented today are not final,” they wrote. “We are still collecting subjects,” and “the patterns we have seen so far could change.”³⁹⁴ In his 1971 book Genetics of Psychopathology, Rosenthal wrote that “the study is still in progress.”³⁹⁵ Because the study’s planned stopping point, assuming that there was one, was known only to the researchers, they were able to use their “hidden flexibility” to continue adding participants in the hope of finding enough spectrum diagnoses among them to produce statistically significant results. Once this was achieved, they had the additional hidden option of stopping the study exactly at that point.

Chambers described behavioral science researchers’ hidden “option to add participants after inspecting the results,” and to “stop data collection as soon as all-important statistical significance is either obtained or seems out of reach.” This can lead researchers to “add participants in an attempt to nudge the p value over the line, without reporting in the published paper that they did so.” Chambers emphasized that “chasing statistical significance by peeking and adding data completely undermines the philosophy of NHST [null hypothesis significance testing].” One “central but often overlooked requirement of NHST,” wrote Chambers, “is that researchers prespecify a stopping rule, which is the final sample size at which data collection must cease.”³⁹⁶ This description of one aspect of biased research practice is very relevant to the critique of the Rosenthal-led study.

The stated purpose of the Rosenthal-led study was to test the components of an “assumed diathesis” (predisposition) of schizophrenia, and one of the study’s four assumptions was that “heredity was an important contributor to schizophrenia.”³⁹⁷ However, in their 1971 report the investigators concluded that “the evidence supports the theory that heredity plays a significant role in the etiology of schizophrenia spectrum disorders.”³⁹⁸ This conclusion is not valid, as a goal of the study was to investigate the components of an already assumed hereditary basis of schizophrenia. The study therefore revolved around the investigators’ circular argument which saw them assume and conclude the very same thing.

In their 1968 publication, Rosenthal and colleagues wrote:

“One inference ought to be drawn from this data: that is, if we are going to learn anything more about the genetics of schizophrenic disorders, we can no longer rely on statistics based only on hospitalized cases. Had we done so in this study, we would have concluded that heredity did not contribute significantly to schizophrenia, or that, if it did, the gene was probably recessive.”³⁹⁹

As this passage suggests, the researchers assumed that schizophrenia was genetically based. Otherwise, they could have “relied” on “statistics based on…hospitalized cases” and simply concluded that their results showed that “heredity did not contribute significantly to schizophrenia,” or that heredity did not contribute to schizophrenia at all. Apparently, these were not acceptable conclusions, compelling the researchers to go beyond hospital diagnoses, and to add additional participants, until they were able to “learn” something about the assumed genetic basis of schizophrenia.

This is not how objective scientific experiments are supposed to be conducted. As stated in the Stanford Encyclopedia of Philosophy, “Scientific objectivity is a characteristic of scientific claims, methods and results. It expresses the idea that the claims, methods and results of science are not, or should not be influenced by particular perspectives, value commitments, community bias or personal interests, to name a few relevant factors.”

The often denied reality in the social and behavioral sciences is closer to psychologist George Albee’s description. Albee concluded in 1982 that his earlier belief that social scientists discover facts in order to build theories was wrong, and that they and others more often “select theories that are consistent with their personal values, attitudes, and prejudices, and then go out into the world, or into the laboratory, to seek facts that validate their beliefs about the world and about human nature, neglecting or denying observations that contradict their personal prejudices.”⁴⁰⁰ This is often referred to as confirmation bias. Another observer wrote, “We ask scientists to perform a virtually impossible feat—to remain resolutely skeptical about the theories they care most passionately about. We shouldn’t be surprised if few scientists can live up to this ideal of behaviour.”⁴⁰¹

Had the Danish-American researchers come close to achieving the “virtually impossible feat” of reaching complete scientific objectivity, they would have accepted and welcomed a conclusion that their studies found no evidence supporting genetic factors, as much as they would have any other conclusion. The problem isn’t that research and confirmation biases exist, but rather that the impact of bias is usually denied or downplayed. The result is that little is done in the social and behavioral sciences to help mitigate it. Although “we may never be able to eliminate bias altogether from human nature,” Chambers wrote, a “sure way to immunize ourselves against its consequences…is peer-reviewed study preregistration.”⁴⁰²

Diagnoses Among Biological Half-Siblings Were Not Consistent with Genetic Predictions

Contrary to genetic predictions, in the Kety-led 1975 study there were more spectrum diagnoses among the biological half-siblings (second-degree relatives, with a 25% average genetic resemblance) versus the biological first-degree relatives (with a 50% average genetic resemblance). Amazingly, 28.5% of the index biological half-sibs received a spectrum diagnosis in the 1975 study.⁴⁰³ In the mid-1970s, some leading schizophrenia researchers pointed to this excess of half-sibling spectrum diagnoses, with Gottesman and Shields noting that “genetic theory predicts a much higher risk for full siblings.”⁴⁰⁴ Benjamin wrote, “This finding is peculiar and contradictory. It shows, in effect, that the less consanguinity, the greater the ‘genetic’ effect” (italics in original).⁴⁰⁵ As Rosenthal once wrote, in order to “demonstrate that genes have anything to do with schizophrenia,” an investigator must show that “the frequency of schizophrenia in relatives of schizophrenics [is] positively correlated with the degree of blood relationship to the schizophrenic index cases.”⁴⁰⁶ The Kety-led studies failed this test.

In the Kety-led studies, the researchers counted spectrum diagnoses among first-degree biological relatives and second-degree biological relatives with equal weighting. When Lidz and Blatt challenged this practice in 1983, Kety responded that he and his colleagues had given “considerable thought” to this question, and that there was a much reason to give the half-siblings “twice the weight” as the full-sibs. They decided in favor of a weighting of one, he wrote, because it “was more likely to be closer to the truth than the extreme values of two or one-half.”⁴⁰⁷ Yet as late as 1972, Rosenthal was writing that when comparing schizophrenia rates among siblings, “we assume that two half-sibs equal one full sib.”⁴⁰⁸

In direct contrast to their schizophrenia studies, in a 1986 Danish-American adoption study of “affective disorders,” Wender, Kety, Rosenthal and colleagues counted half-siblings with one-half the weight of the full-sibs:

“The problem with half-siblings is how to weight them compared with the first-degree relatives (parents and siblings). One approach is to halve the number of half-siblings: the rationale is that a half-sibling is one half as likely as a full sibling to manifest a genetically transmitted disorder. Calculations based on that assumption were also made and will be discussed below.”⁴⁰⁹

As always, the Danish-American researchers’ decision-making “approaches” had a direct impact on their conclusions. In their 1986 affective disorder adoption study, they arbitrarily decided to use a different method of counting half-siblings than they used in their schizophrenia studies.

The “Compelling” Biological Paternal Half-Sibling Comparison Was Not Statistically Significant

The Kety-led 1975 study found a significant concentration of spectrum disorders among the index versus control biological relatives. The researchers viewed this result as “compatible with a genetic transmission for schizophrenia, but it is not entirely conclusive.”⁴¹⁰ Because of possible factors such as “in utero influences, birth trauma, and early mothering experiences,” in the 1975 study Kety and colleagues wrote, “One cannot, therefore, conclude that the high prevalence of schizophrenia illness found in these biological relatives of schizophrenics is genetic in origin” (bold and italics added).⁴¹¹ This clear statement in an original research publication, which Kety repeated in several other publications, did not prevent countless subsequent commentators and textbook authors from concluding that this “high prevalence” did indeed show that schizophrenia is “genetic in origin.” ⁴¹²

Kety and colleagues then made their case for the discovery of “compelling evidence” in support of the “significant operation” of genetic factors, on the basis of a comparison between the biological paternal half-siblings of their index and control adoptees:

“The largest group of relatives which we have is, understandably, the group of biological paternal half-siblings. Now, a biological paternal half-sibling of an index case has some interesting characteristics. He did not share the same uterus or the neonatal mothering experience, or an increased risk in birth trauma with the index case. The only thing they share is the same father and a certain amount of genetic overlap. Therefore, the distribution of schizophrenic illness in the biological paternal half-siblings is of great interest.”⁴¹³

The researchers counted 16 spectrum diagnoses among these paternal half-siblings, but found a “highly unbalanced” diagnostic distribution (14/63 index, versus 2/64 control). They concluded, “We regard this as compelling evidence that genetic factors operate significantly in the transmission of schizophrenia.”⁴¹⁴ Kringlen, on the other hand, believed that the high rate among these half-siblings “is, from a genetic point of view, meaningless,” and that attention should instead be focused on their “life-histories.”⁴¹⁵

It should be noted that Kety and colleagues would have found no statistically significant difference if they had included all schizophrenia spectrum diagnoses in this comparison. In the 1968 and 1975 Kety-led studies, and in the Rosenthal-led studies, the investigators still counted “schizoid or inadequate personality” (Category C) as a schizophrenia spectrum disorder.⁴¹⁶ Nevertheless, they removed Category C diagnoses from their “compelling” 1975 paternal half-sibling comparison. As seen on page 418 of a 1976 Danish-American article published in Schizophrenia Bulletin, based on the researchers’ own calculation the index versus control spectrum diagnosis difference was not statistically significant (p = .094) when spectrum Category C is included.⁴¹⁷

In order to obtain statistically significant results, in their 1975 publication Kety and colleagues excluded spectrum Category C from the paternal half-sibling comparison. This means that the comparison identified by Kety and colleagues as providing “compelling evidence” in support of genetics was not statistically significant according to their own 1975 definition of the schizophrenia spectrum—a fact rarely mentioned in mainstream psychiatry and psychology textbooks.⁴¹⁸

In 1975, Rosenthal defended the inclusion of Category C in his study while at the same time arguing that Category C should not be counted in the Kety-led studies:

“In the Kety study, which we call the Family study, the findings strongly favor the view that diagnostic categories such as uncertain schizophrenia and borderline schizophrenia are genetically related to classical process schizophrenia, but schizoid personality [Category C] is not. Thus, we now talk about a hard spectrum and a soft spectrum… . Undaunted, I want to show you why I think the soft spectrum, which includes a number of syndromes that we call schizoid, is indeed genetically related to process schizophrenia.”⁴¹⁹

And in a 1978 publication, Kety, Rosenthal and colleagues wrote that the question of whether Category C should remain in the spectrum “remains unresolved”:

“The diagnoses of schizoid or inadequate personality were evenly distributed between the biologic index relatives and the biologic control relatives, so this study failed to adduce evidence of a relationship between these syndromes and schizophrenia. Some of our other studies support such a relationship, so the question remains unresolved.”⁴²⁰

In their jointly authored publications of the mid-to-late 1970s, Kety and Rosenthal wrote that although the evidence suggested that Category C was unrelated to chronic schizophrenia, the evidence was not conclusive. This allowed Kety to remove spectrum C diagnoses from the paternal half-sibling comparison (see above), and allowed Rosenthal to retain C diagnoses in the studies he led, permitting both to claim statistical significance for the comparisons they put forward as providing compelling evidence in favor of genetics.

The 1975 paternal half-sibling comparison is an obvious example of data dredging. Committing the “Texas sharpshooter’s fallacy,” the investigators searched through their results until they found a comparison that they believed supported the genetic transmission of schizophrenia, and then concluded that this comparison provided “compelling evidence” in favor of such transmission. The proper method would have been to refrain from “drawing a bullseye” around the “bullet hole” biological paternal half-sibling comparison, but instead to note that the comparison generated an interesting hypothesis that could be tested in a future study. In the subsequent Kety-led 1994 Provincial Study attempt to “replicate” the Copenhagen results, there were 40 identified biological paternal half-siblings of the 29 chronic schizophrenia index adoptees. Not one of these paternal half-siblings was diagnosed with chronic schizophrenia.⁴²¹

In 1976, after the publication of the Kety-led 1968 and 1975 studies, the researchers wrote, “The schizophrenia spectrum was and still is a hypothesis or group of hypotheses on which we hoped our continuing studies might cast some light.”⁴²² There was, however, a clear relationship between Kety and associates’ willingness or unwillingness to remove a diagnosis from the spectrum, or from a particular comparison, and the usefulness of that diagnosis to them in achieving statistically significant results in the genetic direction.

The Final Results of the Rosenthal-Led Study: Negative

As we have seen, in the 1968 and 1971 Rosenthal-led studies the researchers did not arrive at consensus (global) diagnoses for the index and control adoptees, but instead relied on what they called “thumbnail diagnostic formulations” prepared by the Danish psychiatrist co-authors Joseph Welner and Fini Schulsinger. Through personal correspondence with several of the researchers, Lewontin and colleagues discovered that these “thumbnail” formulations did not specify whether an adoptee was in or out of the spectrum, a decision that was “made in a manner and by parties unknown.”⁴²³

In a rarely cited 1978 publication in the Archives of General Psychiatry (now JAMA Psychiatry) by Richard Haier, Rosenthal, and Wender, the researchers finally reported their index and control group adoptee consensus (global) diagnoses. As it turned out, there was no statistically significant difference between these two groups on the basis of an investigator-defined schizophrenia spectrum consisting of chronic schizophrenia, borderline schizophrenia, and “schizophrenic personality” (index 32% versus control 26%).⁴²⁴ The researchers reported these results, but chose to discuss them only in the context of whether scores on the psychological test they administered (the MMPI) differentiated between the two groups.

These glossed-over findings revealed that, based on the Rosenthal-led study’s final definition of what counted as a consensus schizophrenia spectrum disorder, there was no significant schizophrenia spectrum rate difference between the index and control adoptee groups. This should have led the researchers, and all subsequent textbook authors, to conclude that the Rosenthal-led study found no evidence that genetic factors influence schizophrenia or schizophrenia spectrum disorders.⁴²⁵ As Lewontin and colleagues concluded in 1984, “despite the widely misleading early reports on the Rosenthal et al. study, its outcome was in fact negative.”⁴²⁶ I showed in Chapter 5 of The Missing Gene that misleading descriptions of this study, and psychiatry’s refusal to recognize its negative results published in one of its leading journals, have continued into the 21^st century.⁴²⁷ (According to Google Scholar, as of August, 2017 the Rosenthal-led 1971 publication had been cited over 340 times, whereas the Haier et al. article had been cited only 26 times.)

In their subsequent descriptions of the Rosenthal-led study and its findings, the researchers did not mention the Haier et al. paper or its negative results. In his 1983 response to Lidz and Blatt, Kety claimed that in the Rosenthal-led study, “we did not make independent and consensus diagnoses in the ‘adoptee’ studies but merely cited the interviewer’s diagnostic impressions.”⁴²⁸ Other publications failing to mention the Haier et al. results included a 1980 textbook chapter by Rosenthal entitled “Genetic Aspects of Schizophrenia” (one of his last publications), a 1988 chapter in The New Harvard Guide to Psychiatry by Kety and Matthysse entitled “Genetic and Biochemical Aspects of Schizophrenia,” and Wender and Klein’s 1981 book Mind, Mood, and Medicine: A Guide to the New Biopsychiatry.⁴²⁹ In this latter publication Wender ignored his own 1978 Haier et al. consensus spectrum diagnosis results, and overlooked the fact that Rosenthal had to use manic-depressive parents to find statistically significant results in the study he led. Wender and Klein then made their contribution to Danish-American adoption study mythology by claiming that the Rosenthal-led study results showed that “adoptees born of schizophrenic parents and reared by nonschizophrenic parents are indeed more likely to be schizophrenic than adoptees born of nonschizophrenics and reared by nonschizophrenics.”⁴³⁰

Inconsistencies in Counting and Diagnosing Relatives

In the Kety-led studies, the researchers made arbitrary decisions about how to count and diagnose dead or unavailable relatives, usually in ways that helped support genetic interpretations of their data.⁴³¹ The case of the biological father of control adoptees C9, discussed earlier, is one of several examples.

Another documented example was provided by Lewontin and colleagues, who we have seen had been in correspondence with one of the psychiatrists conducting the interviews. Lewontin and colleagues described the case of the biological mother of Copenhagen index adoptee S-11:

“The woman’s mental health records had been edited and then diagnosed blindly by the investigators in 1968. The diagnosis was inadequate personality—at that time, inside the spectrum. The 1975 paper…indicates that, upon personal interview, the woman had been diagnosed as a case of uncertain borderline schizophrenia….But personal correspondence has revealed that the woman was never in fact interviewed; she had committed suicide long before the psychiatrist attempted to locate her, and so—from the original hospital records—she was ‘pseudo interviewed.’ Perhaps the most remarkable aspect of the story…is that the woman had been hospitalized twice—and each time had been diagnosed as manic-depressive by the psychiatrists who actually saw and treated her….We can only marvel at the fact that the American diagnosticians, analyzing abstracts of these same records, were twice able to detect—without ever seeing her—that she really belonged within the shifting boundaries of the spectrum.”⁴³²

Problem Areas in the 1974 Crossfostering Study

In the 1974 Crossfostering Study publication by Wender, Rosenthal, Kety, and colleagues, the researchers produced after-the-fact (post hoc) definitions and combinations that allowed them to conclude that genetic factors play a role in causing schizophrenia, and that being raised by a spectrum-diagnosed parent does not. Because this is the least-cited of the three main Danish-American studies, I will list a few points related to it while noting that this study suffered from many of the problems found in the Kety-led and Rosenthal-led investigations.⁴³³ Problem areas in the 1974 Crossfostering Study include the following points:

· Wender and colleagues moved away from the spectrum concept, and instead used a 1-20 point scale of diagnostic “severity” to rate adoptees in each group

· Although the researchers concluded that “familial psychopathology (as measured by psychiatric diagnosis of the parents) does not” play a role in the “etiology of the schizophrenias,” they produced a table in a different publication showing that 26% of the crossfostered group (CF) adoptees were in the schizophrenia spectrum⁴³⁴

· Fifteen of the 28 CF parents (54%) were diagnosed with “acute schizophrenia,” or as “doubtful schizophrenic or schizoid.” The Danish-American researchers later decided that the acute and schizoid diagnoses did not belong in the schizophrenia spectrum, meaning that they should have subsequently recalculated the results on the basis of a much smaller CF parent group

· The adoptive control (AC) group spectrum diagnosis rate was around 25%, leading the researchers to admit that their AC group “is indeed a nonrepresentative population.”⁴³⁵ About one-third of the AC biological parents were diagnosed with a spectrum disorder, providing additional evidence that, as the researchers put it, theirs was a “poor control group”⁴³⁶

· The researchers used post-data-collection comparisons to support the genetic position

· Statistically significant differences between important comparison groups were not found

· Although unmentioned in the original 1974 publication, the average age of the crossfostered adoptees at the time of their adoptive parent’s first hospitalization was 12-years-old⁴³⁷

By the 1980s, Wender recognized that his Crossfostering Study “was particularly difficult,” and that “the question of what would happen if children born of normal parents were placed in the homes of typical schizophrenics cannot be answered” (italics added).⁴³⁸ Not only was the study “difficult” for the researchers to perform, it is also difficult for readers to follow their tortuous post-data-inspection attempts to get their results to fit expected patterns.

“Uncertain” Diagnoses in the Kety-Led Provincial Study

In the Kety-led Provincial Study, the researchers arbitrarily decided to retain and drop diagnoses as they went along, over a 19-year period. In a normal scientific experiment, researchers define the concept under study (the “dependent variable”) at the outset of their study. Here the dependent variable was “schizophrenia,” which, as we have seen, Kety and colleagues defined as a constantly changing set of “schizophrenia spectrum disorders.” The preliminary Provincial Study findings, based on records, were first published in 1978. The final interview-based Provincial Study publication appeared in 1994. Here I will briefly examine how the researchers defined and re-defined the spectrum during this 16-year period.

The Provincial Study began in 1975. In a 1978 publication, Kety, Rosenthal, Wender and colleagues defined the Provincial spectrum as consisting of “chronic schizophrenia,” “latent schizophrenia,” “acute schizophrenia,” “uncertain schizophrenia,” and “schizoid personality.”⁴³⁹ The Danish members of the team began interviewing Provincial relatives in 1980. In 1983 Kety wrote that, based on the 1975 Copenhagen results, “schizoid and inadequate personality” were “excluded from subsequent analyses” because “there was no…justification for believing that [they]…were related to schizophrenia.”⁴⁴⁰ In a 1985 publication, however, Kety again defined the spectrum as consisting of “chronic, latent, acute, and uncertain schizophrenia schizoid and inadequate personality as described in DSM-II.”⁴⁴¹ In 1988, Kety continued to count Provincial Study relatives diagnosed with one of these conditions as falling within the schizophrenia spectrum.⁴⁴²

By 1992, fourteen years after the publication of the Provincial Study preliminary results, Kety decided to narrow the definition of the schizophrenia spectrum:

“Certain minor modifications evolved in our criteria or in their use between the Copenhagen and the Provincial studies. These included a decreased utilization of the DSM-II diagnoses ‘acute schizophrenia’ or ‘pseudoneurotic schizophrenia,’ probably because of their lack of association with schizophrenia in the earlier study, and a recognition that ‘chronic’ and ‘latent’ schizophrenia were distinguishable whereas the gradations between ‘definite,’ ‘uncertain,’ or ‘probable’ were not.”⁴⁴³

In this passage, Kety wrote of “minor modifications” of spectrum criteria that had “evolved” between the Copenhagen and Provincial Studies—as if researcher decision-making (“degrees of freedom”) had played only a minor role in this “evolution.” Had Kety chosen to describe this process clearly, he would have written that he and his colleagues, with the results in hand and over 14 years into the study, decided to perform a major modification of the spectrum by removing the “acute,” “uncertain,” and “probable” diagnoses. In this 1992 publication, the spectrum was defined as “chronic schizophrenia,” “latent schizophrenia,” and possibly also “schizoid personality.”⁴⁴⁴ (Kety did not clearly define the diagnostic limits of the spectrum in this 1992 report.)

In the final 1994 Provincial Study publication based on interviews, Kety and colleagues defined the spectrum as consisting only of “chronic schizophrenia” and “latent schizophrenia,” even though we saw that in 1978 it had consisted of “chronic schizophrenia,” “latent schizophrenia,” “acute schizophrenia,” “uncertain schizophrenia,” and “schizoid personality.”⁴⁴⁵

There is reason to believe that Kety decided to remove the “uncertain” and “probable” diagnoses in 1992 because their inclusion would have led to results that did not support genetic conclusions. According to Kendler and Diehl, who inspected the Kety-led Provincial Study data while in preparation, “Latent and uncertain schizophrenia was [sic] not found to be significantly more common in the biologic relatives of the schizophrenia adoptees than in those of the control adoptees (6.5% vs. 5.5%, respectively).”⁴⁴⁶

Reduction of the Control Group in the Kety-Led Provincial Study

In 1983, Kety reported that the Provincial Study control group consisted of 42 adoptees.⁴⁴⁷ However, by 1994 there were only 24 control adoptees because Kety decided to remove 18 (43%) of them, almost two decades after the study was initiated. Kety decided that all 13 interviewed control adoptees diagnosed with a “serious or confounding mental illness” (mainly non-spectrum affective disorders) should be removed from the study.⁴⁴⁸ Although a case can be made for removing a control adoptee diagnosed with schizophrenia, from the genetic standpoint, control adoptees diagnosed with an affective disorder would not “confound” the study’s results. Kety decided to remove five additional Provincial control adoptees on the grounds that they refused (or were unable) to be interviewed, even though he had decided to retain 11 non-interviewed control adoptees in his 1975 Copenhagen control group. Kety did not disclose whether he and his colleagues found spectrum disorders among these control adoptees’ biological relatives.⁴⁴⁹

Additional Problem Areas in the 1994 Kety-Led Provincial Study

Other problem areas in the Kety-led 1994 Provincial Study’s attempt to “replicate” the earlier Copenhagen results in the rest of Denmark include the following points: (1) The study was subject to most of the problems and biases found in the 1968 and 1975 studies, and like these earlier studies there was selective placement bias in the sample (see below). (2) In Kendler and colleagues’ 1994 “independent” analysis of the Provincial results based on DSM-III diagnostic criteria, the chronic schizophrenia rate among all index biological relatives of adoptees with the same diagnosis was 2.3% (2/88). This rate was not significantly higher than the control biological relative rate of .6% (1/162), or versus the general population expectation.⁴⁵⁰ This comparison remained statistically non-significant when limited to first-degree relatives.⁴⁵¹ (3) There was no statistically significant first-degree biological relative index group elevation of “latent schizophrenia” when compared with controls.⁴⁵² (4) The symptoms of “latent schizophrenia” were vague, and were often indistinguishable from non-spectrum psychiatric diagnoses. (5) The researchers counted a total of 3 diagnoses of “chronic schizophrenia” among the 24 first-degree biological full-siblings of the 29 index adoptees with the same diagnosis. However, all three were members of the same family, suggesting that family environment, rather than genetic background, best explains this clustering of chronic schizophrenia among three siblings reared together in the same family.⁴⁵³

Summary of the Danish-American Adoption Studies Critique

The basic story of the Danish-American schizophrenia adoption studies is that, after several years of painstaking research in Denmark in the 1960s, the investigators found no evidence that chronic schizophrenia was caused by hereditary factors. Instead of acknowledging this, and then publishing these negative results with little recognition, or the potential for receiving prestigious awards, Kety, Rosenthal, Wender and their Danish colleagues decided to risk opening up what Rosenthal called a “Pandora’s box” by greatly expanding the definition of schizophrenia—Rosenthal admitted that they had “strained” to do so—in an attempt to find results that confirmed their pre-existing belief that genetic factors underlie schizophrenia.⁴⁵⁴ As Bentall concluded, “When Rosenthal and Kety were unable to find clear genetic effects with a conventional definition of schizophrenia they simply expanded their definition….The spectrum concept was merely introduced so that the researchers could get the result they wanted.”⁴⁵⁵

As seen clearly in their published works, the Danish-American researchers:

· Failed to adequately address potentially confounding issues such as prenatal environment, late separation, late placement, range restriction, the unrepresentativeness of the birthparents and the adoptive parents, and the psychologically harmful environments experienced by adoptees between separation and placement

· Used admittedly “vague” diagnostic criteria while expanding the definition of schizophrenia “as widely as it may have ever been reasonably conceived before,” and then narrowed the definition of schizophrenia in the Kety-led Provincial Study

· Retained and rejected spectrum diagnoses as they went along, in a manner consistent with promoting genetic explanations of their data

· Appear to have changed definitions and comparison groups after obtaining and reviewing the data

· Removed spectrum diagnoses from “compelling” index-control comparisons to achieve statistically significant results

· Counted “inadequate personality” and the triply vague “uncertain borderline schizophrenia” diagnoses as schizophrenia spectrum disorders in the Kety-led 1968 and 1975 studies, and in the Rosenthal-led study

· At times counted “manic-depression” as a schizophrenia spectrum disorder, and then denied that they had done so

· Included being gay (“homosexuality”) or “perverse” as a diagnostic indicator of schizophrenia

· Circularly assumed the genetic basis of schizophrenia in the processes of testing for it

· Violated statistical assumptions

· Provided very little information on adoptees’ life histories

· Used arbitrary methods of counting diagnoses and relatives

This illustrates how conclusions drawn from ostensible scientific experiments are transformed into a statement of the investigators’ and their field’s beliefs. Human genetic research has a long history of these types of conclusions, going all the way back to Francis Galton in the 19^th century.

The Danish-American adoption series was a three-decade long exercise in genetically biased deceptive research practices that included p-hacking, HARKing, and data dredging. The researchers were able to reject the null hypothesis, and conclude in favor of genetics, only by employing such practices. The “results” of these massively flawed studies have been endorsed by psychiatry textbooks and review articles written by authoritative experts, and by similar publications in the related fields. In the process, generations of students, clinicians, “patients,” researchers, and others have been greatly misled about the true causes of schizophrenia and psychosis.

Selective Placement: The Achilles’ Heel of Schizophrenia Adoption Research

We have seen that researchers studying families, twins, and adoptees assess the behavioral and diagnostic status of people growing up in environments that they were unable to design, control, or observe, requiring these researchers to make assumptions about these environments. Family, twin, and adoption studies are very far from being controlled scientific experiments, where all factors other than the variable in question (genetics in this case) are strictly controlled.

A major aspect of psychiatric adoption research is the often unstated “no selective placement assumption.” Because children are not randomly assigned to available adoptive homes and then observed as they are growing up, researchers must assume that factors relating to the adoption process (including the policies of adoption agencies) did not lead to the systematic placement of certain types of children into environments contributing to a higher rate of the disorder in question. In the words of Barry Hutchings and Sarnoff Mednick, who performed adoption studies of criminal and anti-social behavior based on Danish samples, “indirect, nongene-mediated relationships between the biological parents’ deviant characteristics and the adoptee’s life experiences must be ruled out before sound interpretations of the findings are possible.”⁴⁵⁶

According to a group of psychiatric genetic researchers, “the critical underlying assumption in adoption studies” is that

“the genetic and environmental variables are separated by the adoption process. This is not true if the trait in question is correlated with an environmental variable for which there is some matching between adoptive and biological parents.”⁴⁵⁷

Schizophrenia adoption researchers must assume that children were not systematically placed into adoptive homes correlated with the socioeconomic status—or presumed genetic status—of their biological families. We will see that in all schizophrenia adoption studies, however, the evidence suggests that experimental (index) group adoptees did experience more psychologically harmful rearing environments than experienced by the control adoptees. Children with a biological family history of mental disorders were seen as inferior potential adoptees, and it is therefore likely that they were placed into more chaotic and psychologically harmful (and potentially more “schizophrenogenic”) adoptive families.

Most adoptees (abandoned children) were placed in the early-to-middle part of the 20th century in Denmark, the United States (Oregon), and Finland. The Swedish study index relatives and adoptees were born between 1932 and 2002. Although rarely mentioned by the original investigators or by the authors of secondary sources, like Germany all four regions had laws permitting the compulsory eugenic sterilization of people labeled “schizophrenic,” “insane,” “feeble-minded,” and so on. Large segments of society, including “scientists” such as Rüdin, Kallmann, and many others, believed that the offspring of people with “insanity” in their family background were the undesirable and dangerous carriers of “hereditary taint.”

Mid-20^th century schizophrenia researcher Manfred Bleuler (the son of Eugen Bleuler) described the “sinister shadow” of “familial tainting” during this era, which obviously played a major role in the adoption process:

“If one knows schizophrenics and their families well, it is sometimes a matter for despair to see how much they suffer under the terrible concept of ‘familial tainting.’ Like a sinister shadow it darkens the lives of many people and of entire families. The stifling, uncertain fear of coming from an ‘inferior breed,’ of carrying within one’s self the seeds of something pathological, morbid, and evil (I am speaking in the jargon the afflicted apply to themselves), like a curse that you must pass on to someone else, causes oppressive feeling of inferiority.”⁴⁵⁸

Denmark

In 1929 Denmark became the first European nation to pass a eugenics-inspired sterilization law.⁴⁵⁹ This law (a stronger law was passed in 1935) was in force until well after the last studied Danish adoptees were placed. (Placements in the Danish-American studies were made between 1924 and 1947.) The forced sterilization of the “insane” and other “eugenically undesirable” people in Denmark continued well into the 1960s.⁴⁶⁰

The Danish adoption agencies checked a potential adoptee’s genetic family background to determine his or her suitability (or desirability) for adoption. As seen in the 1946-47 annual report of the Mother’s Aid Organization of Copenhagen, which was the largest adoption service in the country, the Danish authorities believed that a future adoptee’s genetic “developmental potential” was “of great importance” for the “correct placement” of the child:

“Before a child is cleared for adoption, it is investigated with respect to health, and an attempt is made to obtain detailed information on the child’s family background and to form an impression of its developmental potential. Not only for the adoptive parents, but also for the child itself, these investigations are of great importance for its correct placement. Information is obtained on the child’s mother and father; on whether or not there are serious physical or mental illness in the family background; criminal records are obtained for the biological parents; and in many cases school reports are obtained. By means of personal interview with the mother an impression of her is formed. Where information is uncovered on convicted criminality or on mental retardation, mental illness, etc. in the family background, the case is referred to the Institute of Human Genetics of Copenhagen University, with whom there exists a valuable cooperation for advice on the advisability of adoption” (italics added).⁴⁶¹

When the agency suspected a family history of mental disorders, it consulted the Danish Institute of Human Genetics, which was the keeper of the Danish National Psychiatric Register. The Register was established in 1937, and included the names of people diagnosed with mental disorders, and people who had been psychiatrically hospitalized.⁴⁶² It was created at a time when eugenic ideas and practices were at their high point in Denmark (and in the world as a whole), and was used to aid eugenic selection programs. It is therefore ironic that the Psychiatric Register, which enabled the Danish-American adoption studies to be performed, was created to aid eugenic programs and eugenics-influenced adoption placement decisions.⁴⁶³

According to Mednick, “Every weekend (at least in the 1930s), Danish people who wished to adopt would visit the orphanages and pick children….Children whose selection by an adoptive parent is delayed may be less attractive physically and behaviorally.”⁴⁶⁴ Many children were also perceived as being less attractive genetically. Clearly, the most loving and emotionally stable potential adoptive parents, who were usually informed of “deviance” in the adoptee’s family background, would not have picked children “tainted” by a family history of mental disorders.

Rosenthal understood that knowledge of an index parent’s condition may have influenced adoption agencies’ placement decisions. For this reason, most adoptees in the Rosenthal-led study were placed before their parent was diagnosed with a spectrum disorder, theoretically reducing this potential bias. What the researchers overlooked, however, is that on either environmental or genetic grounds the biological families of an index parent (proband) would be expected to contain a higher number of psychiatrically diagnosed people, and we have seen that Danish adoption agencies checked for “serious mental illness in the family background.” It is therefore likely that the adopted-away children of Danish birthparents who developed a psychotic disorder were placed into inferior rearing environments compared with the control children, regardless of whether their birthparent was or was not diagnosed at the time of adoption. This also holds true for the adoptees in the Tienari-led Finnish study.

Kety believed that his “Adoptees’ Family” design was less vulnerable to selective placement factors because it began with diagnosed adoptees, in contrast to the studies that began with diagnosed parents.⁴⁶⁵ However, although he and his colleagues did not mention this, in 8 of the 33 Copenhagen index adoptive (IA) families, but in none of the 34 Copenhagen control adoptive (CA) families, a parent had been admitted to a mental hospital at some point.⁴⁶⁶ This lends additional support to the position that the Danish-American adoptees were placed into more psychologically unstable adoptive homes than were the control adoptees.

Oregon

Similar conditions existed in Oregon, where adoptees were placed between 1915 and 1945. Although Heston made no mention of it, Oregon passed a law in 1917 creating a “State Board of Eugenics,” whose duty was to authorize the compulsory sterilization of “all feeble-minded, insane, epileptic, habitual criminals, moral degenerates and sexual perverts,” because they might produce “inferior” offspring.⁴⁶⁷ An additional 1919 Oregon law stipulated that if a person had been admitted to a mental hospital, this constituted “prima facie evidence that procreation by any such person would produce children with an inherited tendency to feeble-mindedness, insanity, epilepsy, criminality or degeneracy.”⁴⁶⁸ The author of a 1925 article in Eugenical News wrote approvingly that, while sterilization laws in many U.S. states were limited to “inmates of institutions,” in Oregon “there is a Eugenics Commissioner, who has authority to comb the state for degenerates and enforce sterilization.”⁴⁶⁹ Because Heston’s index adoptees were born to women hospitalized with schizophrenia, it is very unlikely that the “tainted” children they gave birth to were placed into the same types of adoptive homes as were the “non-tainted” control children.

In 2002, Oregon Governor John Kitzhaber officially apologized for the “great wrong done to more than 2,600 Oregonians over a period of about 60 years,” who endured “forced sterilization in accordance with a doctrine called eugenics.” He added that “most of these Oregonians were patients in state-run institutions. The majority of them suffered from mental disorders and disabilities.”⁴⁷⁰

A 1929 article by American eugenicist Paul Popenoe captured the thinking of the times. He believed, in general, that American adoptees’ “ancestry” was not “up to par,” although “the best are sorted out early by the child-placing agencies. The remainder, collected in orphanages, represent predominantly the inferior levels and usually show up badly in tests.” Indeed, 25 of the 47 Heston index adoptees had spent months or years “collected in orphanages.”⁴⁷¹ Popenoe advised potential adoptive parents to “pick out a child with as good ancestry as possible,” which in “the mother is always known.”⁴⁷² Clearly, the “inferior” biological offspring of “insane” mothers were not seen as desirable potential adoptees. As Kringlen once observed in relation to Heston’s study, “Because the adoptive parents evidently received information about the child’s biological parents, one might wonder who would adopt such a child.”⁴⁷³

Finland and Sweden

In 1935 the Finnish Parliament passed the Sterilization Act, which allowed the compulsory eugenic sterilization and castration of “idiots,” “imbeciles,” and the “insane,” which included people carrying the schizophrenia and manic-depression labels.⁴⁷⁴ In 1950, Finland passed the Castration Act, which permitted the castration of criminals, the mentally retarded, and the “permanently mentally ill.” Compulsory eugenic sterilization was not legally abolished in Finland until 1970. As we have seen, around one-third of Tienari’s index adoptees were placed after their mother was diagnosed/labeled as psychotic.⁴⁷⁵ Sweden also had a long history of eugenics and compulsory eugenic sterilization.⁴⁷⁶

Conclusion

It appears that selective placement factors were operating in all schizophrenia adoption studies, which indicates that these studies’ “critical underlying assumption” was violated. As behavioral genetics pioneer Gerald McClearn (1927-2017) wrote about adoption studies in a 1964 work, in an era when genetic researchers in the behavioral arena wrote much more cautiously and defensively about their claims, “Insofar as selective placement exists, of course, there is a genotype-environment correlation which makes the accurate assessment of the relative contributions of heredity and environment impossible.”⁴⁷⁷ The selective (non-random) placement of adoptees was a major confounding factor in these studies, and we have seen that they were subject to several other confounding environmental factors as well.

Misleading Textbook Accounts of Schizophrenia Adoption Research

In 2002, Jonathan Leo and I reviewed Nobel Prize winning psychiatrist Eric Kandel’s chapter on schizophrenia in the 2000 edition of Principles of Neural Science. We highlighted some errors, and noted the genetic/biological bias in Kandel’s discussion of schizophrenia genetic research.⁴⁷⁸ In Chapter 5 of The Missing Gene, I documented major problems with psychiatry and abnormal psychology textbook accounts of schizophrenia adoption research. These problems include that their authors frequently (1) emphasized the original researchers’ conclusions at the expense of independent critical analysis; (2) relied on secondary sources; (3) failed to discuss, or mentioned only briefly, the views and publications of critics; (4) misreported studies’ methods and results; (5) failed to discuss problems with the reliability and validity of a schizophrenia diagnosis in the context of genetic research; (6) failed to discuss potential adoption study environmental confounds such as prenatal environments, late separation, late placement, range restriction, and selective placement; (7) cited studies failing to find statistically significant results in the genetic direction as evidence in favor of genetics; and (8) accepted the original researchers’ definition of schizophrenia without question, and used the word “schizophrenia” when discussing what were in fact Kety, Rosenthal, and Wender defined “schizophrenia spectrum disorders.” Unfortunately, little has changed since 2006. In Chapter 10 of The Missing Gene, I showed that a parallel process has occurred with adoption studies of bipolar disorder.⁴⁷⁹

I will now elaborate on the above point # 4, and introduce an additional area of textbook misreporting. The main publications I refer to are among the leading textbooks in psychiatry: The APA’s prestigious American Psychiatric Publishing Textbook of Psychiatry, edited by Robert Hales and colleagues, and the Kaplan and Sadock textbooks. The latter series comes in three main varieties: the large two-volume Kaplan & Sadock’s Comprehensive Textbook of Psychiatry, the single-volume Synopsis of Psychiatry: Behavioral Sciences/Clinical Psychiatry, and the Concise Textbook of Clinical Psychiatry. These textbooks have been in use for decades, with new editions appearing every few years.

Citing Non-Existent Studies in Support of Genetics

We have seen that apart from the handful of anecdotal single-case reports that I discussed briefly in Chapter 4, reared-apart schizophrenia twin studies do not exist. Nevertheless, in the 1996 edition of Kaplan and Sadock’s Concise Textbook of Clinical Psychiatry, we find the following description:

“Monozygotic twins have the highest concordance rate. The studies of adopted monozygotic twins show that twins who are reared by adoptive parents have schizophrenia at the same rate as their twin siblings raised by their biological parents. That finding suggests that the genetic influence outweighs the environmental influence.”⁴⁸⁰

Although it is slightly possible that Kaplan and Sadock had in mind the single-case anecdotal reports, their readers likely came away with the false impression that systematic genetic studies of schizophrenia, using reared-apart MZ twins, had been performed. Kaplan and Sadock described studies using reared-apart MZ pairs, where one group consisted of MZ twins raised by their biological parents, while the other group consisted of their adopted-away (separated) MZ co-twins raised by their adoptive parents. They reported that these alleged studies showed that both twin groups have the same schizophrenia rate, leading to the conclusion that the “genetic influence outweighs the environmental influence.”

The above unreferenced discussion of non-existent “studies of adopted monozygotic twins” was repeated in Kaplan and Sadock’s 2004 edition of the Concise Textbook, but was removed by the 2008 edition. It was also repeated in the 1998 and 2003 editions of Synopsis of Psychiatry, but was removed by the 2007 Synopsis edition.⁴⁸¹

Misleading accounts are also found in Hales and colleagues’ American Psychiatric Publishing Textbook of Psychiatry. In a passage found in the 2003 edition’s “Genetics” chapter by James Knowles, which was repeated in the 2008 edition chapter by Prabhakara Choudary and Knowles, it was written that “two studies of MZ twins reared apart have shown high pairwise concordance for schizophrenia, providing further evidence for a genetic component in the etiology of this disorder.”⁴⁸² Again, no reared-apart twin studies for schizophrenia exist. The authors cited a secondary source, Gottesman and Shields’s 1982 book Schizophrenia: The Epigenetic Puzzle, in support of their claim. This book was a review and analysis of the causes of schizophrenia, but provided no new data. In their discussion of individual reared-apart MZ twin pairs, Gottesman and Shields counted a grand total of 12-14 such pairs in the “worldwide literature.” Their discussion of the “two studies” cited by Choudary and Knowles related to a pair of investigations conducted in Japan. However, Gottesman and Shields wrote that they had previously “erred” in counting these studies because of their “overlap,” and because “most of the Japanese pairs were reunited before age 5.” They wrote that counting these two studies constituted “a source of error in the literature on such pairs.” Choudary and Knowles thereby transformed a “source of error” into “further evidence for a genetic component.” Once again, unknowing readers are misled. Gottesman and Shields concluded that, while researchers can learn from the stories of reared-apart MZ twins, “the data are not the kinds one would choose to use in building a solid foundation for either a genetical or an environmental theory about the etiology of schizophrenia.” ⁴⁸³

Calling on Gottesman’s work to help expose these errors is ironic, given that an underlying theme of my first two books and other writings (see in particular The Missing Gene, Chapter 6) was a documentation of his five decades of omitting or twisting key historical facts and scientific findings, ignoring or misrepresenting the views of critics, and largely bypassing the unwinnable twin method EEA debate, in support of psychiatric genetic and behavioral genetic positions.⁴⁸⁴ Unfortunately, other leaders of these fields have used similar tactics for decades.

The authors of a chapter on schizophrenia in the Hales-edited 2011 edition of Essentials of Psychiatry wrote, “Twin adoption studies have been remarkably consistent in reporting approximately 50% concordance rate [sic] for monozygotic twins.”⁴⁸⁵ This is the standard claim for studies of reared-together pairs (see Chapter 4), but we have seen that a “twin adoption” study is another name for a reared-apart twin study.

In summary, leading psychiatry textbooks have reported that genetic theories of schizophrenia are supported by seemingly convincing studies of reared-apart MZ twin pairs, even though no such studies exist. This has occurred in the bipolar disorder arena as well, where the 1999 U.S. Report of the Surgeon General claimed that “in studies of monozygotic twins reared separately (‘adopted away’), the results also revealed an increased risk of depression and bipolar disorder compared with controls.” The two studies cited by the Report were adoption studies that did not use twins.⁴⁸⁶

Incorrect Descriptions of the Comparison Groups Used in the Kety-Led Studies

We have seen that in the Kety-led studies, conclusions in favor of genetics were based mainly on comparisons between the index biological relative (IB) and control biological relative (CB) groups (see Figure 5). Kety, as we saw, believed that comparing index biological relatives (IB) to index adoptive relatives (IA) was “difficult,” “inappropriate,” “improper,” and “fallacious.” Nevertheless, Kety appeared to cite the “fallacious” IB versus IA comparison in 1974 when it served his and psychiatry’s purposes. At other times, Kety compared the IB diagnostic rate to the combined rate in among the IA, CB, and CA relative groups. Still, the main comparison Kety and colleagues cited in support of genetic conclusions was IB versus CB for relatives in general, and for the 1975 paternal half-siblings in particular. In their final Provincial Study publication, which combined the Copenhagen and Provincial results to produce the Danish “National Sample,” they clearly stated that their conclusions were based on IB-CB comparisons: “In the combined ‘National Sample’ of adoptees with chronic schizophrenia, that disorder was found exclusively in their biological relatives [IB] and its prevalence overall was 10 times greater than that in the biological relatives of controls [CB].”⁴⁸⁷

In order to provide additional evidence that many textbook authors, authors of popular works, and influential psychiatric genetic researchers do not know the original adoption study publications as well as they should, here I cite a few instances where these authors described the Kety-led studies’ comparison groups incorrectly.

Michael Egan and Thomas Hyde contributed the “Schizophrenia: Neurobiology” chapter to the 2000 edition of Kaplan & Sadock’s Comprehensive Textbook of Psychiatry. After citing family and twin studies they wrote, “Of seven adoption studies, all found an increased incidence of schizophrenia in biological relatives [IB], but not in adoptive relatives [IA]. This data convincingly demonstrates that genetic factors rather than shared, familial environmental factors are at work.”⁴⁸⁸ Due to their design, the Heston, Rosenthal-led, and Tienari-led studies could not have made such a comparison, and in the Kety-led studies the researchers instead chose to compare IB versus CB relatives. In bestselling author Matt Ridley’s 2004 book The Agile Gene: How Nature Turns on Nurture, he wrote that Kety “found that schizophrenia was 10 times as common in the biological relatives of diagnosed schizophrenics who had been adopted as children as it was in their adopting families.”⁴⁸⁹ An additional mistaken claim that the Kety-led studies were based on IB-IA comparisons is found in the 2008 edition Kaplan and Sadock’s Concise Textbook.⁴⁹⁰

Similarly incorrect claims that the Kety-led studies’ conclusions were based on IB versus IA comparisons are found in the publications of many of the world’s leading psychiatric genetic researchers. These works include Tsuang and Vandermey’s 1980 Genes and the Mind; a 1991 chapter by Michael Lyons, Kendler, Gersony Provet, and Tsuang in Genetic Issues in Psychosocial Epidemiology; Jerrold Maxmen and Nicholas Ward’s 1995 Essential Psychopathology and Its Treatment; Brien Riley, Philip Asherson, and Peter McGuffin’s “Genetics and Schizophrenia” chapter in the 2003 textbook Schizophrenia; the Cardno and McGuffin, and the Owen, O’Donovan, and Gottesman chapters in the 2003 book Psychiatric Genetics and Genomics; and a 2005 “Genetics of Schizophrenia” article by Patrick Sullivan.⁴⁹¹

In schizophrenia researcher Lynn DeLisi’s 2017 popular work 100 Questions & Answers About Schizophrenia: Painful Minds, she wrote that in the Kety-led 1968 study and the Rosenthal-led study, the “investigators showed that an excess of schizophrenia was present in the biological relatives of individuals with schizophrenia [IB], but not the adoptive relatives [IA].”⁴⁹² Again, Kety and colleagues made no such comparison in 1968. Like Ridley and countless other authors, DeLisi misled the public when she wrote these studies counted diagnoses of “schizophrenia,” when in fact they counted spectrum disorders ranging all the way down to “uncertain borderline schizophrenia,” “inadequate personality,” and “manic-depressive psychosis.”

In a 1999 book published by the American Psychiatric Press on genetic research, psychiatrist Joel Paris wrote that the Kety-led 1975 study showed “that if an adopted child has a schizophrenic biological parent, his or her risk for the disorder increases.”⁴⁹³ This is the common claim made about the Rosenthal-led study, but the design of the Kety-led study could lead to no such finding, since the “proband” index and control adoptees were chosen on the basis of whether or not they were diagnosed with a spectrum disorder. And in fact, only one of the 66 biological parents of the index adoptees was diagnosed with chronic (B1) schizophrenia in the 1975 Kety-led study.⁴⁹⁴ A similar error is found in Nancy Andreasen’s 2001 Brave New Brain, where like Paris she mistakenly described Kety’s study as using the groups and group comparisons used in the Rosenthal-led Adoptees Study.⁴⁹⁵

A Past President of the APA Gets It Wrong

Former APA President Jeffrey Lieberman wrote a 2016 op-ed piece in the Wall Street Journal, where he also cited research that never took place in support of genetic theories of schizophrenia.⁴⁹⁶ Lieberman is the author of over 500 papers and articles published in the scientific literature, and has written and/or edited at least 16 books on mental disorders and psychiatry, including at least two dealing specifically with schizophrenia. He was the lead editor of the 2006 American Psychiatric Publishing Textbook of Schizophrenia.⁴⁹⁷ When he speaks and writes, he does so with the great authority of a widely published leading schizophrenia expert.

Lieberman and his Wall Street Journal co-author Ogi Ogas claimed that Kety’s “data” produced a “genetic riddle,” that the MZ co-twin of someone diagnosed with schizophrenia is “50% likely to develop the illness—not the 100% expected.” Kety, however, did not study twins, nor did he publish any original schizophrenia twin data.

Again referring to “Kety’s data,” Lieberman and Ogas wrote that “if both your parents had schizophrenia, you were only 50% likely to develop the illness.” Here they were apparently referring to so-called “dual mating” studies of the offspring of two parents diagnosed with schizophrenia. These studies have their own set of problem areas, and the rates are actually lower than 50%.⁴⁹⁸ But Kety never performed a dual mating study either.

Like many others, in their 2015 book Shrinks: The Untold Story of Psychiatry, Lieberman and Ogas incorrectly described how the Kety-led study “settled the question of schizophrenia’s genetic basis”:

“Kety started a new study. He identified individuals with schizophrenia who had been adopted at birth and examined the rates of schizophrenia among both adoptive [IA] and biological [IB] relatives. He found higher rates of schizophrenia in the biological relatives [IB], but not in the adoptive families [IA]….These findings demonstrated that schizophrenia was at least partially due to one’s genetic endowment.”⁴⁹⁹

We have seen that Kety’s conclusions were based on a different group comparison, and that this comparison was based on schizophrenia spectrum disorders, not “schizophrenia.” In addition, Kety did not claim that the adoptees in his study were “adopted at birth” (although about half were removed from their families within the first month of life).⁵⁰⁰

Lieberman and Ogas, like many other textbook authors, seemed unfamiliar with the original “evidence” that psychiatry ceaselessly puts forward as proof that its disorders are both valid and “highly heritable.” They and others paint a picture of airtight studies, when in reality these studies leak like sieves.

***

For many of the influential authors discussed in this section, the genetic basis of schizophrenia is so established as fact that it doesn’t seem necessary to carefully check the original sources and studies they cite. Why bother, when the debate was “closed” decades ago. It follows that few psychiatrists or behavioral scientists question the twin method’s all-important equal environment assumption, or read through the often tedious original schizophrenia adoption study publications. In the process, they provide at times stunningly inaccurate accounts of the original research and its meaning.

The misleading, incomplete, or false accounts of schizophrenia adoption research by textbook authors and the authoritative leaders of psychiatry and psychiatric genetics that I have documented here and elsewhere are just the tip of a disturbing iceberg of misinformation, and exemplify the conclusion I reached in The Missing Gene that textbooks have played a huge role in perpetuating the mythology around schizophrenia genetic research.⁵⁰¹ “In general,” I wrote in 2006, “the surveyed textbooks have rubber stamped the original investigators’ and contemporary psychiatry’s conclusions about the results of schizophrenia adoption research. Their descriptions are frequently inaccurate, and leave the general impression that their authors did not carefully review the original studies, and sometimes did not even read them.”⁵⁰² Only four of the 43 textbooks I surveyed in 2006 cited a publication critical of a schizophrenia adoption study’s methods and conclusions, and only three provided their own limited critical analysis. And it has only gotten worse since then.

Schizophrenia Adoption Research: Summary and Conclusions

In this chapter I have taken readers on a journey through the most frequently cited adoption studies that have been systematically presented in standard textbooks in a misleading manner since the 1970s. These texts have ignored or concealed methodological, factual, and even scientific-ethical flaws that, in addition to twin studies based on the false assumption that MZ and DZ environments are equal, call into question everything that we supposedly “know” about the genetics of schizophrenia.⁵⁰³

Earlier in this chapter I summarized the major invalidating problems of schizophrenia adoption research. I focused on the Danish-American studies because of their groundbreaking nature, and because they are the most frequently cited schizophrenia adoption studies. In the words of a leading schizophrenia researcher quoted at the beginning of this chapter, these studies played a major role in helping turn “the corner for support for ‘nurture’ to support for primarily ‘nature.’” I also highlighted some major problems with the Heston and Tienari investigations, and alerted interested readers to publications containing the full critique of these studies.

Like many critics over the decades, I challenged the claim that the adoption process itself makes a clean separation between genetic and environmental influences on schizophrenia. Due to confounding factors such as selective placement, prenatal environment, late separation, range restriction, unrepresentativeness, and late placement, the evidence strongly argues against this claim. Environmental confounds are glossed over or are unmentioned in most psychiatry textbooks, and readers usually are told nothing about the social conditions and eugenic policies in the societies in which adoptees (abandoned children) grew up.

Another important issue was the obvious genetic bias of the researchers themselves. Kety, Rosenthal, and Wender arbitrarily retained, rejected, and combined diagnoses as they went along on the basis of whether or not these diagnoses were “genetically related to chronic schizophrenia,” thereby creating a self-fulfilling prophecy that produced results that led to the desired genetic conclusions. As Stephen Jay Gould showed in The Mismeasure of Man, human genetic research has a long history of investigators “shifting criteria to work through good data toward desired conclusions,” and creating conditions in which “data” are not allowed to “overthrow…assumptions.”⁵⁰⁴ As a careful examination of their published works reveals, the Danish-American researchers continued this unfortunate tradition—simply refusing to permit their data to overthrow psychiatry’s need to establish “schizophrenia” as a valid genetic/biological medical disorder. The broad term now used to describe this practice is p-hacking, which includes various forms of scientific manipulation, misconduct, or fraud. When false results produced by p-hacked research have great social, scientific, and political importance, and affect or harm the lives of millions of people while entire fields look on, it constitutes a scientific scandal.

The methodological problems, the ever-changing definition of “schizophrenia spectrum disorders,” the after-the-fact data manipulations, the group comparison changes, the statistically non-significant results, and the numerous environmental confounds (biases) I have described in this chapter lead to the conclusion that the Danish-American adoption studies have completely failed to provide scientifically acceptable evidence in favor of genetics. Genetic interpretations of their results must therefore be rejected outright. In a just world, textbooks would be rewritten to reflect this conclusion, and their authors would remove these studies from the “causes of schizophrenia” section and would discuss them in a section documenting misconduct in science.