Bonnie Cleaveland, PhD ABPP
“Much of what we know about other people is based on the stories they tell us about themselves. Unfortunately, sometimes what they tell us is not true.” – Newman & Baumeister (1996)
“[V]arious debunked mental health theories continue to exert inappropriate influence over the decisions of family courts.” – Nichols (2013)
Some research is purely theoretical. It doesn’t have an immediate impact on people. Other research, however, immediately impacts the peoples’ lives. We all agree we want what’s best for children. To do that, we must base our legal decisions on the best possible science, not on conjecture. In the case I’m going to talk about, the research affects children who are relying on the court systems to advocate for their best interests.
The Split Feather pilot study (Locust, 1994; Locust, 1988; Locust, 2000) is frequently cited in child welfare cases involving American Indian children in court and was even appended in toto to a Supreme Court of the Unites States ruling (Brief for Tanana Chiefs Conference, et al., 2013). Unfortunately, the study was implemented so poorly that we cannot draw conclusions from it.
Twenty American Indian (AI) adoptees, removed as children from their biological families by child welfare agencies and placed with non-native families, were interviewed as adults. The study’s author concluded that removal from their tribes created the “split feather syndrome” (SFS) in which removal from one’s American Indian culture caused distress and a host of psychological problems.
Split feather may or may not be a valid theory. It’s impossible to know, because the existing research (one pilot study) is inadequate to study the concept. Sadly, because many judges and attorneys, and even some caseworkers and other professionals, are not familiar with the research, results that may be very wrong are leading to the wrong outcomes for children.
Let’s start from the very beginning. Why is methodologically sound research important?
Although we like to believe humans are rational, we have numerous, systematic errors in our thinking (Kahneman, 2013).
For example, there has been a great deal of controversy about whether vaccinations cause autism. Vaccines prevent disabling and fatal diseases, and failing to vaccinate causes potential harm to the individual and to society. The scientist, Andrew Wakefield, who “discovered” a link between autism and vaccines faked his data (Godlee et al., 2011). There is no relationship between vaccines and autism. Even when the fraudulent science was exposed, and when the scientific community weighed in with other research, many people didn’t change their beliefs in line with reality. Scientists, however, are trained to overcome the natural human tendency to maintain pre-existing beliefs despite the evidence.
I detail some of the cognitive biases, below.
We have limited and biased perspective. For example, how do you know the earth is spherical and not flat? Sitting in one place, the earth certainly looks flat. It’s only when you get a bigger perspective, seeing pictures of the spherical earth from space, that it becomes obvious that the earth is not flat.
I heard someone say after the 2008 election of Barack Obama, “How did he possibly get elected? I don’t know a single person who voted for him!” The flaw in the logic is obvious. It’s what’s called “anecdotal evidence.” This person was from an area of the country that’s strongly Republican. She was likely to have similar political views as her family and friends, since we learn from our families and we group ourselves with others who are similar to us. Republicans naturally gravitate to other Republicans because their values are similar. It can be absolutely true that she doesn’t know anyone who says they voted for Barack Obama. But she also didn’t know a random sample of people. Further, there is a percentage of people who simply won’t express their views to her.
For this reason, researchers try to randomly sample people in ways that don’t bias the research. Gallup, for example, randomly chooses landlines and cellphone numbers throughout the 50 states and Washington, D.C. for their surveys in the U.S.A. They sample throughout the country, rather than in one part of the country, so as not to be biased by geography. The sample adults of all ages, and they interview in Spanish for Spanish-speaking respondents.
If we want to know what snack a specific third-grade child’s class would prefer, we might simply ask each of the 25 children and tally the results. Since the population is small, we can survey them all and know what the child’s class wants.
We wouldn’t be sure, though, if the results are generalizable to all classes of third graders throughout the United States. And we’re fairly certain it would NOT generalize to all 8 and 9 year-old children throughout the world.
If we wanted to know the preferred snack for all 3rd graders in the United States, we couldn’t ask them all. Luckily, a sample of 3rd graders is sufficient. Obviously, there are regional variations in our food choices. We can’t sample 3rd graders in Texas and expect it to generalize to 3rd graders throughout the United States. Only a representative sample will adequately represent the population of third graders in the United States. That would mean sampling 3rd graders throughout all states.
The sample size has to be large enough to represent the national population of third graders. If we sampled only 100 third graders, even with a random sample, we’re likely to get some answers that are not frequent. We might find that 10% say broccoli is their favorite food. But if we sample a large enough sample, the infrequent answers are a smaller and smaller percentage of the total. The sample represents the population better as the sample gets larger.
The more variation in the survey responses, the larger sample is needed to be sure of the results. If the population of children overwhelmingly lists only five favorite foods, a smaller sample is needed than if they have 100 favorite foods. A sample size calculator and article by The Survey System explains the factors that influence how large a sample is needed to accurately represent the population.
We are likely to come to conclusions based on our immediate experiences, even though they don’t represent the full range of experiences. This systematic cognitive error is called the availability heuristic. In order to minimize such bias, one must study a sample of the entire group to which the researcher wishes to generalize her results.
Let’s say you’re a physician who treats only the most difficult, recurrent cancers. In fact, the other oncologists in town refer their most difficult cases to you. Because you get the most difficult cases, a large percentage of your patients die. If you relied only on your own experiences, you might believe that most people who are diagnosed with cancer die from their cancer. In a situation in which not much data is available, we’re more likely to rely on our personal experiences. If our personal experiences don’t represent the population (and they probably don’t), our conclusions are wrong. Luckily, there are good scientific data on survival rates of various cancers, and, as an oncologist, you’re likely to be aware of the data.
Another error our brains make is giving too much weight to authority. We are overly confident in information presented by people of high status or whom we trust, a logical fallacy called “Appeal to Authority.” In a related bias, it’s much easier to adopt an idea when the majority believes it (Esser, 1998; Wang, 2001) even though the number of people believing in an idea has no relation to its truth.
Humans are quite susceptible to confirmation bias – an often unconscious process that’s difficult to resist (Powell, 2012). In confirmation bias, we accept information that fits with our theory and dismiss disconfirming evidence. Although the process of science is objective, scientists must fight against bias and limited perspective (Medin & Bang, 2014), especially in qualitative research.
Scientists are trained to look for evidence that does not confirm their theories.
Touching a hot stove is the simplest of experiments. Because the effect follows the cause quickly, it’s fairly easy to learn what happens when you touch a hot stove with bare skin. But how should you treat a minor burn? It’s more difficult to understanding cause and effect when the effect is immediate. When I was a child, we put ice on burns. Intuitively, ice makes sense. But it’s completely wrong.
Research shows that ice doesn’t help a burn and it can cause tissue damage (Cuttle, Kempf, Kravchuck et al., 2008; Cuttle & Kimble, 2010). Instead, cool running water decreases pain and swelling, decreases damage on the cellular level, improves healing, and decreases scar formation. How do we know? Researchers treat standardized burns are treated with one treatment (ice) vs. another (cool running water). Variables such as wound depth, size and temperature are measured in the two groups. The two groups must then be significantly different – not just slightly different – to determine that one treatment is better than the other.
When effect is much later than its cause, or when there are multiple causes or multiple effects, it’s very difficult to find out what works without an experiment. Grandma never found out about cool running water vs. ice, because Grandma didn’t have a comparison (control) group. We need science to answer complex questions.
It’s also difficult to find out why something happens without an experiment. Why does burned skin blister? Why does cool running water work better than ice? If we can figure out why something happens, we have a chance to develop interventions to change it. On the other hand, if we misattribute effect to the wrong cause, we’re not likely to find interventions that work to fix the problem or to prevent it in the first place.
Human beings suffer from illusory correlation; we are surprisingly bad at determining cause and effect (Tversky & Kahneman, 1974).
Educated and intelligent people are sometimes led astray by the normal human cognitive fallibilities. As a result, humankind has evolved a scientific method to test and refine our theories. Science is a systematic method, not an outcome. It’s based on logic and critical thinking. Science minimizes natural human bias. Constructs (ideas being studied) must be measurable, and they must be replicable, so that the results found in one lab can be also tested in another, unrelated lab. Any single study could have major errors or fraud. Scientists report their methods and share their data. If other researchers don’t get similar results, the results are thrown out.
A scientific experiment is never perfect, but it strives for the best knowledge available.
The study examining the Split Feather Syndrome suffers from many such cognitive errors and has not been tested with the scientific method.
Dr. Carol Locust interviewed 20 American Indians who were adopted or fostered into white culture as children. She said that “every Indian child” is at risk of “long-term psychological damage” as a result of out of culture placement (Locust, 1998; (Locust, 2000; Native Canadian, n.d.). Further, “19 out of 20 have psychological problems related to their placement in non-Indian homes.” She is clearly stating that the removal caused psychological problems. Here are her hypotheses:
Split Feather Hypotheses
Hypothesis 1: Being taken from one’s culture causes impaired identity.
Hypotheses 2: Being taken from their culture causes impaired identity for American Indians greater than other children taken from their culture.
Hypothesis 3: Impaired identity causes problems such as substance abuse and mental illness.
These hypotheses may or may not be true. However, we can look to prior scientific results to make more educated hypotheses:
Alternative Hypothesis 1: Being taken from one’s family causes trauma (Ahrens, Garrison & Courtney, 2014).
Alternative Hypothesis 2: Being taken from one’s family unfairly (without cause) causes additional trauma over and above Alternative Hypothesis 1.
Alternative Hypothesis 3: Abuse and neglect causes trauma (Felitti, Anda, Nordenberg et al., 1998; Spinnazola, et al., 2014).
Alternative Hypothesis 4: Trauma is associated with increased rates of substance abuse and mental illness (Brady & Back, 2012).
Alternative Hypothesis 5: Being a visible minority is likely to lead to discrimination; another form of trauma (Becares, Nazroo, & Stafford, 2009; Jones & Galliher, 2014)).
Here, we have two sets of hypotheses. We can’t know which are correct without research. All but one of the alternative hypotheses above are supported by extensive, existing research. On the other hand, the primary Split Feather hypotheses are untested because the study was wholly inadequate.
Although she didn’t publish detailed methods for the study, it seems pretty clear that Locust had a biased sample. Dr. Carol Locust opened her files, collected over time, in which American Indians who were adopted or fostered outside of their culture approached her or answered her advertisement. She found through interviewing them that many of them were dissatisfied about their adoptions. It’s likely that her sample was biased by the fact that Locust became known as a researcher studying Native American adoption distress, and then used at least some people who approached her as subjects of her study.
It’s easy to imagine that these two ads would attract vastly different samples:
“Are you a Native American harmed by your adoption into a non-Indian home? We are studying the harmful effects of adoption.”
“Are you a Native American who is adopted? We are seeking subjects for an academic study of adoption.”
Unfortunately, Locust didn’t disclose her specific recruitment efforts in detail. I Typically, a researcher makes public the actual text of any subject recruitment advertisements.
In a true experiment to find out whether removal from home and culture would negatively affect children, experimenters would randomly assign children to be removed from their homes or not. A control group is similar in all possible ways to the experimental group. One way this is accomplished is by randomly assigning subjects to the control group (those not removed) or to the experimental group (those removed). For ethical reasons, we wouldn’t do such a study.
There are two solutions. First, a researcher could use a control group, such as AI adoptees adopted within AI culture. Second, a researcher could use a statistical control group.
Consider a thought experiment – one we would never do in reality. Imagine we flipped a coin to remove children from their homes vs. keep them in the home. How is the psychosocial functioning for the children removed from their homes vs. those not removed? We would find a certain level of mental health problems in both groups. Here’s what we really want to know: Is the level of mental health problems and substance abuse and identity problems greater in the group of children removed from their homes? This tells us only about removal vs. non-removal under unfair conditions (random removal).
The split feather concept alleges that out of culture removal is more destructive for American Indian children than for other children. Ergo, the study requires large groups of American Indian children removed and not removed, and large groups of non-Native children removed and not removed (also into a different culture). In our thought experiment, then, we would have AI and non-AI children removed randomly or not. The control group is essential to tease apart the effects of removal versus the effects of AI political affiliation on removal trauma.
In psychosocial research there are many questions for which a true random controlled experimental design is unethical. So researchers attempt to eliminate as much bias as they can with ethical research studies. In this case, for example, American Indians living with their tribes would make a good control group. Non-native adoptees could be a second good control group. Also, researchers can statistically control for things they can’t control for experimentally. For example, since we know that traumatic experiences affect substance abuse and mental health (Enlow, Blood & Egeland, 2013), we might statistically control for the effects of trauma in our study.
The SF study didn’t have a control group of any kind. As a result, we have no idea if AI children adopted out of culture have worse outcomes than AI children adopted within their culture.
Factors other than out-of-culture placement are very likely to cause poor psychosocial outcomes. Additional factors might make negative outcomes more likely or less likely among children of adoption. The following factors should be measured in any study of adoptees, but they were not measured in the SF study.
Child abuse and neglect are known to have devastating and lasting affects on children’s mental and physical health (Felitti, et al. 1998, Spinazzola et al 2014). Youth in foster care experience rates of witnessed domestic violence, abuse, neglect, andparental substance abuse higher than the general population (Ahrens et al., 2014).
We can presume that at least some percentage of the SF adults removed from their home had experienced abuse or neglect in their family of origin (prior to removal). The split feather pilot study doesn’t control for (or even mention) abuse or neglect the children suffered prior to removal. Locust does note that some children were abused or neglected in their adoptive families. Even if none of the children were abused in their within-culture homes, the abuse by adoptive families (rather than removal from culture) could completely account for their psychological distress in adulthood.
Locust mentions that discrimination is a factor that contributes to split feather. Discrimination creates negative psychological impacts (Whitbeck et al., 2001). Discrimination is almost certainly more pervasive and pronounced for people who are minorities within the wider culture than for minorities living amongst their culture. (Becares, et al., 2009).
Minorities experience discrimination even when living within their culture. American Indian children are likely to have experienced discrimination when they were living with their family and tribe. Discrimination, then, is a factor that needed to measured and considered in the data.
It’s likely that children removed as infants are less likely to be affected than children at older ages, especially those removed during the pre-teen and teen years (Sharma, McGue & Benson, 1996, Part I; Sharma, McGue & Benson, 1996, Part II).
A child removed from a home with very serious sexual, physical or emotional abuse is likely to experience relief (as well as stress) with removal, whereas a child unfairly removed from a loving, stable home is likely to have significant difficulty with the unfairness of the removal.
Studies have shown that open adoptions are healthy for children and appreciated by birth and adoptive families (e.g., Berge, Mendenhall, Wrobel & McRoy, 2006). In some cases of severe and continuing abuse, severing ties with the abuser is obviously better for a child. However, even in those cases, maintaining contact with extended family and siblings may provide stability for a child in out of home care.
Compared to children not in out of home care, children in foster care have higher levels of health and mental health problems prior to their removal (Ahrens et al, 2014). A child who is abused or neglected prior to removal or who has severe emotional problems prior to removal is likely to fare worse in post-removal placements.
Lack of supportive adults and friends in a child’s environment is likely to have a negative affect on a child’s adjustment (Edwards & Benson, 2010).
Humans’ ability to control aspects of their situation affects their physical and mental well-being (Kay, Whitson, Gaucher & Galinsky, 2009). Thus, the child’s inability to influence the situation (for example, state her preferences) is likely to worsen the child’s well-being.
In science, it’s very important to describe the study carefully so that it can be evaluated and replicated. Most studies are published in scientific journals, and there’s a standardized way to report study methods.
The split feather study was not published in a scientific journal but in newsletters (Locust, 1994; Locust, 1998; Locust, 2000) and several websites that didn’t contain the standard descriptions of subjects, data, and methods. I asked Locust for her data in a letter (Cleaveland, 2014 August 22). She responded by email saying that her data was confidential (Locust, 2014). Next, I sent her three emails asking for a description of her methods (Cleaveland, 2014). She failed to respond to any of the three emails.
The study was a pilot study, which means that the results are not meant to be seen as reliable or valid or to generalize to the entire population under study. Further research is required. Although Locust described her study as a pilot study at the time (Locust, 1998; Locust, 2000), she later insisted the study was not a pilot study (Locust, 2014, September 23, Personal Communication). A Lexis Nexis academic on August 3, 2014 for the term “split feather” revealed no further research.
Sixteen articles, however, cite the Split Feather Syndrome. This means that this concept, which is not scientifically validated, has been cited in other authors’ work. The SF study was appended, in toto, to a brief to the Supreme Court of the US (Brief for Tanana Chiefs Conference, et al., 2013).
Qualitative research is a great way to start to understand a topic. The Split Feather pilot study is an example of such qualitative research. Qualitative studies have few subjects but study them in depth. Qualitative studies generate hypotheses, but they don’t come to broad conclusions about the whole population.
Elsevier, a large publisher of many scientific journals, describes peer review beautifully. “Peer review is an essential dividing line for judging what is scientific and what is speculation.” When student researchers are learning to research, typically at the masters and doctoral levels, a committee of professors is convened to review the research design before the study is started. In peer review, the journal editor chooses people who are familiar with the kind of research they’re doing. The committee of peers evaluates the research before publication. They’re likely to point out problems in the methods and measures and reject studies that don’t show what they purport to. Peer review assures us that methodological flaws are minimized.
Scientific research relies on collaboration. At numerous steps in the process, others are involved. A funding agency may review the proposal for research. An Institutional Review Board often reviews research for ethics and safety. Then, when the research is considered for publication, an editor reviews it and forms a committee of experts (peer review) to critique the research. Finally, if it passes all of those tests, the research is published in a professional journal where it’s available to a wider audience of scientists and the public. This collaboration provides diverse views and critiques. None of this happened with the SF pilot study.
Research procedures should minimize confirmation bias, including using open-ended and not leading questions. A structured interview, with predetermined questions (rather than questions asked spontaneously), should probably be used in research that will generate hypotheses about the population under study. Short of that, however, in exploratory research, the methods at the very least should be described in order to understand any sources of bias.
In studies that use interviews, it’s normal for the interview questions to be available for review because how you’ve worded a question is likely to influence how people answer. In studies with open-ended questions, it’s very important to administer the questions the same way each time. When questions are published, a study can be replicated. The National Survey of American Life, is a great example (Jackson, Neighbors, Nesse, Trierweiler, & Torres, 2004).
Replication is important in science. When researchers find similar results time after time, the certainty of the conclusions is increased.
Any study that relies on human memory has inherent limitations. Although we all think we remember certain things very clearly, studies show human memory is quite fallible. Humans are subject to false memories (e.g., Berkowitz, Laney, Morris, Garry & Loftus, 2008). People often tend to report things in a way that makes them look better, or in ways they believe the researcher wants them to respond (Tran, Stieger & Voracek, 2012). Finally, most of our narratives about ourselves are likely to be seriously flawed, especially considering the complex factors leading to emotional and physical illness (Keenan, Gallup and Falk, 2003).
To minimize self-report bias, researchers often use other sources of data, such as family members’ reports, medical or legal records, child welfare records, and so on.
Researchers create standardized measures of concepts and they actually research the reliability and validity of the measure itself. Only when you have a reliable and valid measure can you have a good research study. A study typically uses standardized measures, previously determined to have reliability and validity. The SF author didn’t use any standardized measures.
Measures must have two important characteristics to be used in science. First, they should be reliable. For example, blood pressure is a widely used and valuable indicator in medicine. When done in a standardized fashion, blood pressure measurements are reliable. Measuring blood pressure without taking into account arm position, type of chair, bladder fullness, recent nicotine use, talking or listening patient, cuff size, and cuff over clothing creates unreliability in blood pressure measurements (Handler, 2009). If you measure blood pressure improperly, blood pressure is not likely to be a useful predictor of health, because it doesn’t reflect your true blood pressure. Measures used in a study must be reliable, and the reliability of the measures is reported in the paper.
A measure might be reliable but not valid. In that case, it’s not an adequate measure. We can, for example, reliably measure the length of your left thumb. But (at least so far) that measure doesn’t predict anything at all. It’s not a valid measure of, say, your cardiovascular risk.
For example, one website reports: “All 20 of the Split Feathers felt that the [sic] were average or above in intelligence, but half of them had spent time in education remedial programs in school” (Native Canadian, n.d.). We know how to measure intelligence. People cannot reliably report their own intelligence (Gignac, Stough, & Loukomitis, 2004). The primary measure of intelligence used for adults, the WAIS-IV, is quite reliable and predicts academic performance (Wechsler, 2008). It has been validated in IA populations (Nakano, 2013). There’s no reason not to actually measure intelligence if an author wanted to see if, in fact, the subjects had academic performance lower than would be expected given their intelligence. But Locust simply didn’t measure intelligence.
In fact, there weren’t any standardized measures in the split feather study. It appears that Locust observed that AI adoptees had substance abuse and mental health problems (and her observations may have been wrong) and, without evidence, concluded that adoption caused the problems. She failed to account for other experiences that might have led to such problems. Here are some of those potential factors.
As many as 80% of youth in long-term foster care have developmental, physical health and mental health problems (Bellamy , Gopalan & Traube, 2010). Courtney Dworsky, Brown, Cary, Love & Vorhies 2011) studied 732 former foster youth (all ethnicities) at age 23 to 24 in the Midwest Study. Twenty four percent were convicted of a crime after leaving foster care. Twenty one percent had housing instability, homelessness, or were in jail. Twenty-one percent earned less than a high-school degree, and 52% were unemployed.
The suicide rate is two and a half times greater in American Indian and Alaska Native peoples (AI/AN) ages 15-34 than in the general population, and poverty is more than twice as high as in the general population (Gray & McCullagh, 2014).
AI/AN individuals have higher rates of child sexual abuse, substance abuse, domestic abuse, rape, Post Traumatic Stress Disorder and childhood behavioral problems than the general population (Gray & McCullagh, 2014). The decreased access to medical and mental health treatment among AI/AN populations compounds the problem (Cunningham, 1993; Flores & Tomany-Korman, 2008).
Locust asserts that out of culture removal causes substance abuse and psychiatric problems. However, she uses no control group. She doesn’t acknowledge the high rates of trauma, psychiatric and substance abuse among AI/AN people who remain in their culture and among the population of foster children.
These high rates of psychosocial problems could easily account for all of the symptoms Locust found in her subjects.
Age of Removal
It’s likely that children removed from their culture at different developmental stages might be differently affected (Sharma et al., 1996). Any foster care or adoption study should measure and report age at removal. However, the SF study didn’t.
It’s difficult for an abused child to be taken from her school and family, despite her relief at being rescued from an abusive situation. American Indian children were taken in great numbers by social services from their families and tribes in the 1960’s. Some were taken as a result of abuse and neglect. Many, however, were taken in a cultural genocide – with the goal of forcibly acculturating AI children to white culture and wiping out AI culture (Mhyra, 2011). The trauma those non-neglected and non-abused children faced being unfairly ripped out of their loving families and tribes is unimaginable (Sinclair, 2007). The unfairness of removal from a stable family is likely to create a trauma unto itself. American Indians who are aware of their history, as a group, have a history of cultural trauma whether they’re adopted out of culture or remain within their culture. Any adoption or foster care study should certainly measure abuse and trauma in the original AI/AN home and in the non AI/AN foster or adoptive home.
Statistical methods control for confounding variables
When we can’t do a randomized, controlled study, we have statistical methods to do tease apart the relevant factors. Instead of literally separating children into groups that we remove versus don’t remove, we look at very large groups of children and use statistics to assign groups that don’t differ except on the one factor we want to study. You could think of statistically controlling a variable in this way: You decide to play golf with your sister, but she’s a much better golfer. It would be very frustrating for you and not very challenging for her without a handicap. A handicap statistically controls for how good a golfer you are, so you start with a level playing field.
To understand “split feather”, we might study a large group of native and non-native children in the care of the Department of Children’s Services in foster care. It would be ideal to have a measure of their mental health before removal. We would statistically create two groups (native and non-native) with similar pre-existing mental health, age, and trauma history. Then, with those groups being as equal as we could make them, we can see the effect of removal on mental health, substance abuse, and trust. It’s not perfect, but, with a large enough sample of children, it’ll tell us what we need to know.
An Academic Search Complete database search for “split feather” found no results. An author search for “Locust, C” yielded one result, which wasn’t a study of split feather but a paper that mentions the SF construct.
Increasingly, although it wasn’t true when Locust completed her study, researchers now frequently provide raw data from their studies so that other researchers can analyze it. Data is provided without identifying information to protect participants’ privacy, and their consent is obtained before the study. At the very least, however, the methods (for example, the questions used and the protocols used in the study) should be provided.
Locust didn’t quantify the variables and analyze them mathematically. She simply assumed that since AI adoptees had high rates of substance abuse and mental illness, the adoption must have caused the problems. Correlation, however, does not necessarily mean that one variable causes the other.
The US per-capita consumption of cheese over a ten-year period is correlated with death by bedsheet entanglement. It’s a large correlation of .95 – (a correlation ranges from 0-1). Do you believe that eating cheese causes people to die by bedsheet? Correlation is not causation. These two variables probably occur together randomly.
Variables might even occur together non-randomly. In a commonly-cited example, ice cream sales are correlated with drowning deaths. Does ice cream cause drowning? No. It’s more likely that people buy more ice cream and swim more during higher temperatures. George Mason University has a great website called Stats Simplified that further explains the difference between correlation and causation.
Giving something a name makes it seem real (Gasque, 2001; Hardin & Banaji, 1993; Hickman, 2000). But giving something a name does not make it real.
In the late 17th century in Salem, Massachusetts, a number of people were accused of being witches. Their syndrome included such symptoms as pain, fever, convulsions and psychosis. The villagers decided that their symptoms were a syndrome of witchcraft, and many people were put to death. No doubt the symptoms were real to those experiencing them, but the name “witch” falsely attributed the cause to witchcraft.
The Split Feather Syndrome is named, so it seems real. But there’s no evidence for the construct. Anyone can formulate a syndrome, but it doesn’t make it real or useful. And it doesn’t mean the “cause” is that which is explained.
A syndrome has specific, measurable symptoms that frequently occur together. For example, metabolic syndrome contains the following symptoms: obesity, high triglycerides, low HDL cholesterol, high blood pressure and high fasting blood sugar. Calling something a syndrome doesn’t mean there’s a single cause.
Second, a syndrome should be specific. A large-scale study looked at the correlation between seven kinds of trauma in childhood and the top ten causes of illness and death. The study found a strong dose-response effect: the more childhood trauma, the more likely the adult is to have risk factors for the top causes of death such as obesity, substance abuse, suicide attempts, depression, smoking, a high number of sexual partners, and physical inactivity (Felitti et al., 1998). It doesn’t make sense to call the same group of symptoms “Split Feather Syndrome” because the child is American Indian.
Third, a syndrome should be useful and predictive. For example, metabolic syndrome is a useful construct because it predicts type II diabetes and cardiovascular disease (Liu et al., 2014). SFS, by definition, implies causality. Specifically, it implies that taking an American Indian from her tribe causes a feeling of lost identity leading to depression and substance abuse. There’s no study at this time that could come to that conclusion.
There are numerous problems with the concept of split-feather. Yet the split feather construct has been frequently used in family and tribal court over 15 years since its publication. A search of Google Scholar, searching “split feather” in all courts only returned only one result, so it doesn’t seem to be used in appeals and Supreme Courts, where the construct is more likely to be challenged by expert witnesses. It’s likely that the concept has been so ingrained in the culture that SF is simply assumed.
It’s past time for researchers to do a study designed to detect the effects of out of culture adoption on AI children.
The most important aspect of any study is its control group. Since few Caucasian children are adopted into non-Caucasian homes, a comparison group of African American (AA) children adopted by non-AA families might make sense. Of course, we can look at all adoptees, but we want to make sure the primary comparison group has similar characteristics. Although distinct, African Americans have been victims of enslavement and chronic discrimination. This gives us four groups:
|A||American Indian children adopted into a different culture|
|B||American Indian children adopted into the same culture|
|C||Non-American Indian children adopted into a different culture|
|D||Non-American Indian children adopted into the same culture|
For each variable, we’ll look at the difference between those adopted within their culture and those adopted outside their culture. Since most of the measures (e.g., depression, substance abuse, loss of identity) are of undesirable outcomes, we’ll assume that a higher score means poorer functioning.
|Adopted into….||American Indian (NA)||Non-American Indian (NON)|
|Same culture||Minus B||Minus D|
|= AI-O (American Indian Outcome)||= NON-O (Non-American Indian Outcome)|
Let’s call the American Indian Outcome (AI-O) the difference between AI children adopted into a different culture vs. a their own culture.
We’ll call the Non-American Indian Outcome (NON-O) the difference between non-AI children adopted into a different culture vs. a their own culture.
The Split Feather Hypothesis
American Indians adopted into a different culture have poorer outcomes than American Indians adopted into the same culture (A-B>0).
The American Indian Outcome is worse than the Non-American Indian difference between non-native adoptees’ adopted into the same versus a different culture, or AI-O > NON-O.
Thus, the SF Hypothesis is that A-B is worse than C-D for each measure.
Ideally, subjects should be obtained as randomly as possible from the entire population of Native children in foster care / adoption and Non-Native children in foster care/adoption. Of course, resource and practical limitations sometimes dictate a less comprehensive sampling.
It’s possible that existing databases have sufficient information to explore the split feather hypothesis. Research with existing databases, although it has limitations, is a much less resource-intensive proposition. An existing database that may have the necessary data is The Midwest Evaluation of the Adult Functioning of Former Foster Youth. This sample of 596 (at last follow-up) can be compared to a sample from the National Longitudinal Study of Adolescent Health, as was done in a 2014 study of health outcomes for adults formerly in foster care (Ahrens et al., 2014). Ahrens and her colleagues report that the Institute of Medicine made former foster youth a federal research funding priority, so funds may be readily available for an academic researcher.
The National Data Archive on Child Abuse and Neglect (NDACAN) at Cornell University is another source of data for children in out of home care. It’s unclear if the numbers of AI children in the study are large enough for powerful comparisons. The Administration on Children, Youth and Families, U.S. Department of Health and Human Services began the survey of children in the child welfare system in 1999. 5501 children who entered the CWS from October, 1999 to December 2001 have been followed four months after they were first involved with CWS and then at 1 year, 1.5 years, 3 years, and 6-7 years later. However, the database is designed with measures meant to be comparable with other databases, and it may serve as a good comparison along with other data.
A longitudinal study such as the Midwest study would work the best. A study might look at children over several time periods, including at the time they’re placed in foster care, 5 years later, and 15 years after placement. Measuring the children when they’ve just entered foster care gives the opportunity to easily look back to their circumstances and functioning prior to their placement. If children placed in foster care from ages 6 through 16 were studied, then the youngest subject at 15-year follow up would be 21 years old and the oldest would be 31. Unfortunately, only five of the 596 subjects at the age of 26 identified as “American Indian,” leaving an inadequate sample at the oldest age of follow-up so far (Courtney, et. al., 2011).
Where possible, a study should use measures also used in other large-scale studies of child well-being. Such measures are reliable and valid. Further, by using the same measures as in these large studies, some comparisons to other populations are possible.
The Adverse Child Experiences (ACE) Study (Felitti, et al, 1998) provides a good model for studies of childhood events and their effects on medical and psychological outcomes. The National Survey of American Life is another large-scale study that might provide good measures.
A standardized measure of identity should be used as well as a specific measure of AI/AN identity.
Self-report measures have limitations. Medical records, social service records, and reports of collaterals such as teachers, biological and foster parents, and extended relatives provide additional data.
A statistician can compute the number of subjects necessary in each group to have enough power to detect a significant effect.
An initial exploration of currently available data should be the first step.
If out of culture AI adoptees have the same outcomes as in culture-AI adoptees, (with enough statistical power) the SF hypothesis is disconfirmed. Similarly, if out of culture AI adoptees have similar outcomes to out of culture non-AI adoptees (controlling for the higher base rates of many of the outcome variables among AI people), the SF Hypothesis is disconfirmed.
These outcomes disconfirm the SF Hypothesis
A ≤ B
AI-O ≤ NON -O
This outcome confirms the SF Hypothesis
AI-O > NON-O
The Administration for Children and Families has funding opportunities for research listed by agency, including the Children’s Bureau, Family and Youth Services Bureau and the Administration for American Indians. Other potential sources of funding include the Gates Foundation, the Stuart Foundation, The Walter S. Johnson Foundation, the Annie E. Casey Foundation, Casey Family Programs, the Jim Casey Youth Opportunities Initiative, and the W.T. Grant Foundation.
A person may think of herself as a split feather, an addictive personality (Nathan, 1988), or an alien abductee (Newman & Baumeister, 1996). But there’s no evidence for any of those constructs, so they should not be used in courts of law.
The split feather syndrome may exist and be just as Dr. Locust has presented it. It may be completely invalid, or the truth may be somewhere in between. Until we study it properly, however, we cannot know.
Thanks to Nicole Ford and Dr. Darleen Opfer, Rusty Harrison & Eleanor Perkins for examples and consultation in the preparation of this article. Thanks to Rusty Harrison for consultation on research ideas.
ICWA, Indian Child Welfare Act, American Indian, Native American, AI, AN,
Alaska Native, Department of Children’s Services, Department of Children’s Services, DSS, Department of Social Services, Adoptive Couple v. Baby Girl, Adoption, Foster Care, Family, Child Welfare, University of Arizona, Research Methods, Thesis, Dissertation, Adoption, Foster Care, Split Feather Syndrome, Scientific Method, Research Design, National Indian Child Welfare Association, NICWA
Ahrens, K.R., Garrison, M.M., and Courtney, M.E. (2014) . Health Outcomes in Young Adults From Foster Care and Economically Diverse Backgrounds Pediatrics; originally published online November 3, 2014; DOI: 10.1542/peds.2014-1150
Bagley, C. & Young, L. (1980). The long-term adjustment of a sample of inter-country adopted children. International Social Work. 23, 16-22.
Becares, L., Nazroo, J. and Stafford, M. (2009). The Buffering Effects of Ethnic Density on Experienced Racism and Health. Health & Place 15(3) 700-708.
Bellamy, J.L., Gopalan, G. & Traube, D.E. (2010) A National Study of the Impact of Outpatient Mental Health Services for Children in Long Term Foster Care. Clinical & Child Psychology and Psychiatry. 15(4), 467–479.
Berge, J., Mendenhall, T., Wrobel, G., Grotevant, H., & McRoy, R. (2006). Adolescents’ feelings about openness in adoption: implications for adoption agencies. Child Welfare, 85(6), 1011-1039.
Berkowitz, S. R., Laney, C., Morris, E. K., Garry, M., & Loftus, E. F. (2008). Pluto Behaving Badly: False Beliefs and Their Consequences. The American Journal of Psychology, 121(4). 643-660.
Brady, K. T., & Back, S. E. (2012). Childhood trauma, posttraumatic stress disorder, and alcohol dependence. Alcohol Research: Current Reviews, 34(4), 408-413.
Brief for Tanana Chiefs Conference. et al. as Amici Curiae Supporting Respondents, Adoptive Couple v. Baby Girl, 133 S.Ct. 2552 (2013) (no. 12-399).
Cleaveland, B. (2014, August 22). Letter to Carol Locust, Ed.D
Cleaveland, B. (2014, September 24). Email to Carol Locust, Ed.D
Cleaveland, B. (2014, September 29). Email to Carol Locust, Ed.D
Cleaveland, B. (2014, October 3). Email to Carol Locust, Ed.D
Courtney, M.E., Dworsky, A., Brown, A., Cary, C., Love, K., & Vorhies, V. (2011). Midwest evaluation of the Adult Functioning of Former Foster Youth: Outcomes at Age 26. Chapin Hall: University of Chicago. Retrieved from http://www.chapinhall.org/sites/default/files/Midwest%20Evaluation_Report_4_10_12.pdf.
Cunningham, P. J. (1993). Access to care in the Indian Health Service. Health Affairs, 12(3), 224-233.
Cuttle, L., & Kimble, R. (2010). First aid treatment of injuries. Wound Practice & Research, 18(1), 6.
Cuttle, L., Kempf, M., Kravchuk, O., Phillips, G. E., Mill, J., Wang, X. Q., & Kimble, R. M. (2008). The optimal temperature of first aid treatment for partial thickness burn injuries. Wound repair and regeneration, 16(5), 626-634.
Edwards, O.W. & Benson, N.F. (2010) A Four-Factor Social Support Model to Mediate Stressors Experienced by Children Raised by Grandparents, Journal of Applied School Psychology, 26(1), 54-69, DOI: 10.1080/15377900903368862
Enlow MB, Blood E, & Egeland B. (2013). Sociodemographic risk, developmental competence, and PTSD symptoms in young children exposed to interpersonal trauma in early life. Journal of Traumatic Stress, 26(6): 686-94.
Esser, J. K. (1998). Alive and Well after 25 Years: A Review of Groupthink Research. Organizational Behavior & Human Decision Processes, 73(2/3), 116-141.
Felitti, V. J., Anda, R. F., Nordenberg, D., Williamson, D. F., Spitz, A. M., Edwards, V., … & Marks, J. S. (1998). Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults: The Adverse Childhood Experiences (ACE) Study. American journal of preventive medicine, 14(4), 245-258.
Flores, G., & Tomany-Korman, S. C. (2008). Racial and ethnic disparities in medical and dental health, access to care, and use of services in US children. Pediatrics, 121(2), e286-e298.
Gasque, T.J. (2001). The Power of Naming. Vermillion: College of Arts & Sciences, University of South Dakota.
Gignac, G. E., Stough, C., & Loukomitis, S. (2004). Openness, intelligence, and self-report intelligence. Intelligence, 32(2), 133-143.
Godlee, F., Smith, J., & Marcovitch, H. (2011). Wakefield’s article linking MMR vaccine and autism was fraudulent. BMJ, 342.
Gray, J. S., & McCullagh, J. A. (2014). Suicide in Indian country: The continuing epidemic in rural American Indian communities. Journal Of Rural Mental Health, 38(2), 79-86. doi:10.1037/rmh0000017
Handler, J. (2009). The importance of accurate blood pressure measurement. The Permanente Journal, 13(3), 51.
Hardin, C., & Banaji, M. R. (1993). The influence of language on thought. Social Cognition, 11(3), 277-308.
Hickmann, M. (2000). Linguistic relativity and linguistic determinism: some new directions. Linguistics, 38(2), 409.
Jackson, J. S., Neighbors, H. W., Nesse, R. M., Trierweiler, S. J., & Torres, M. (2004). Methodological Innovations in the National Survey of American Life. International Journal of Methods in Psychiatric Research, 13(4): 289-298.
Jones, M. L., & Galliher, R. V. (2014, August 4). Daily Racial Microaggressions and Ethnic Identification Among American Indian Young Adults. Cultural Diversity and Ethnic Minority Psychology. Advance online publication. http://dx.doi.org/10.1037/a0037537
Kahneman, D. (2011). Thinking, fast and slow. Macmillan.
Kay, A. C., Whitson, J. A., Gaucher, D., & Galinsky, A. D. (2009). Compensatory control achieving order through the mind, our institutions, and the heavens. Current Directions in Psychological Science, 18(5), 264-268.
Keenan, J. P., Gallup, G. C., & Falk, D. (2003). The face in the mirror: The search for the origins of consciousness. HarperCollins Publishers.
Liu, L., Miura, K., Fujiyoshi, A., Kadota, A., Miyagawa, N., Nakamura, Y., … & Ueshima, H. (2014). Impact of metabolic syndrome on the risk of cardiovascular disease mortality in the United States and in Japan. The American journal of cardiology, 113(1), 84-89.
Locust, C. (1994, March). Split feathers: Adult American Indians who were adopted into non-Indian homes as children. Presentation made at the Arizona State Indian Child Welfare Conference, Tucson.
Locust, C. (1998). Split feathers study. Pathways. September / October 13(4).
Locust, C. (2000). Split feathers: Adult American Indians who were placed in non-Indian families as children. Ontario Association of Children’s Aid Society, 44(3), 11-16.
Locust, C. (2014, September 23). Personal Communication via email.
Locust, C. (n.d.). American Indian Adoptees. Split Feather Study. Retrieved from http://splitfeathers.blogspot.com/p/split-feathers-study-by-carol-locust.html
Locust, C. (n.d.) SPLIT FEATHERS: ADULT AMERICAN INDIANS WHO WERE PLACED IN NON-INDIAN FAMILIES AS CHILDREN. Retrieved from http://www.nativecanadian.ca/Native_Reflections/split_feather_syndrome.htm
Locust, C., & Lang. (1996). Walking in two worlds: Native Americans and the VR system. American Rehabilitation, 22(2), 2-11.
Medin, D. L., & Bang, M. (2014). The cultural side of science communication. PNAS Proceedings Of The National Academy Of Sciences Of The United States of America, 111(Suppl 4), 13621-13626. doi:10.1073/pnas.1317510111
Mhyra, L. (2011) “It runs in the family”: Intergenerational transmission of historical trauma among urban American Indians and Alaska Natives in culturally specific sobriety maintenance programs. AI and AN Mental Health Research 18(2) 17-39.
Nakano, S. W. (2013). Factor structure of the Wechsler Intelligence Scales for Children – Fourth Edition among referred Native American students. Psychology In The Schools, 50(10), 957-968.
Nathan, P. The addictive personality is the behavior of the addict. Journal of Consulting and Clinical Psychology, Vol 56(2), Apr 1988, 183-188.)
Native Canadian, n.d. http://www.nativecanadian.ca/Native_Reflections/split_feather_syndrome.htm
Newman, LS & Baumeister, RF. (1996). Toward an explanation of the UFO abduction phenomenon: Hypnotic elaboration, extraterrestrial sadomasochism, and spurious memories. – Psychological Inquiry (7)2.
Nichols, A. M. (2013). Toward a Child-Centered Approach to Evaluating Claims of Alienation in High-Conflict Custody Disputes. Mich. L. Rev., 112, 663.
Powell, M. J. (2012). Skill in Interviewing Reduces Confirmation Bias. Journal Of Investigative Psychology & Offender Profiling, 9(2), 126-134.
Sharma, A. R., McGue, M. K., & Benson, P. L. (1996). The emotional and behavioral adjustment of United States adopted adolescents: Part I. An overview. Children and Youth Services Review, 18(1), 83-100.
Sharma, A. R., McGue, M. K., & Benson, P. L. (1996). The emotional and behavioral adjustment of United States adopted adolescents: Part II. Age at adoption. Children and Youth Services Review, 18(1), 101-114.
Sinclair, R. (2007). Identity lost and found: Lessons from the sixties scoop. First Peoples Child & Family Review, 3(1), 65-82.
Spinazzola, J., Hodgdon, H., Liang, L. J., Ford, J. D., Layne, C. M., Pynoos, R., … & Kisiel, C. (2014). Unseen wounds: The contribution of psychological maltreatment to child and adolescent mental health and risk outcomes. Psychological Trauma:
Tran, U.S., Stieger, S. & Voracek, M. (2012). Psychometric analysis of Stober’s Social Desirability Scale (SDS-17): An item response theory perspective. Psychological Reports: Measures and Statistics, 111(3), 870-884.
Tversky, A. and Kahneman, D. (1974). “Judgment under Uncertainty: Heuristics and Biases”. Science (American Association for the Advancement of Science) 185(4157): 1124–1131.
Wechsler D. (2008). Wechsler Adult Intelligence Scale–Fourth Edition. Pearson; San Antonio, TX.
Wang (2011, May 3). Under the Influence: How the Group Changes What We Think. Researchers Study What Gives Social Norms Their Power. Wall Street Journal. Retrieved from http://www.wsj.com/articles/SB10001424052748704436004576298962165925364
Whitbeck, L. B., Hoyt, D. R., McMorris, B. J., Chen, X., & Stubben, J. D. (2001). Perceived discrimination and early substance abuse among American Indian children. Journal of health and social behavior, 405-424.
 Of course, before we had the technology to actually see that the earth is roughly spherical, Aristotle and others hypothesized a sphere based on systematic observations.
 The format of the files is unknown.
 They would then statistically examine whether the two groups were, in fact, the same prior to removal. Were the ages of children in the two groups similar? Were their abuse backgrounds and mental health similar prior to removal?
 Along with some negative effects.
 Undoubtedly, some of the youth in that sample were in out of culture placements.