Thursday, June 27, 2013

Applying David Levy's errors in critical thinking to suicidology


APPLYING DAVID LEVY’S ERRORS IN CRITICAL THINKING TO SUICIDOLOGY
 
David Lester

            Levy (2010) wrote a provocative book on common errors in psychological research and theorizing. He did not focus on suicide at all but, as I read his book, I could see how his errors occasionally creep into our research and theorizing about suicide. Hence this essay in which I explore Levy’s errors in critical thinking in suicidology.

Conceptualizing Phenomena
 
Error 1: Language is Evaluative

            Although language may sound non-judgmental, there is often an evaluative connotation to particular words. Language, then, can affect our thoughts and attitudes and, vice versa, our thoughts and attitudes can affect our language. It is important, therefore, to not present our value judgments as objective reflections of truth.

            A good example here is from Robins, Murphy, Wilkinson, Gassner and Kayes (1959). They looked at the past history of 134 completed suicides and found that only eight were not psychiatrically ill and, of these eight, five had a terminal illness.[1] Are the results of this study objective truths? First, as we will see in Error 2, psychiatric diagnoses are constructs and not real things. Second, a perusal of other studies on this issue finds that the proportion of suicides judged retrospectively to be psychiatrically disturbed ranges from about 33% to the 94% reported by Robins, et al. Thus, it seems that the article by Robins is an opinion. Robins and his team believe that one must be psychiatrically disturbed in order to die by suicide. Their “scientific” study is merely a subjective opinion of normality versus abnormality. Their language is a value judgment that “suicidal individuals are crazy.”

Error 2: The Reification Error

            This refers to the error of regarding abstracts concepts as if they are concrete objects. Levy gives the example of self-esteem. Self-esteem is not a thing that a person has; it is a concept that psychologists have created to explain behavior. In suicidology, we typically study a behavior, attempted or completed suicide, but most of our explanatory concepts are concepts. Levy draws attention to the often-made distinction between physical and mental problems. Physical things are concrete things, whereas mental things are abstract concepts. The unconscious was not discovered; it was invented. Psychosis is not detected; it is declared.

            Levy noted that theories can be event theories (Type E) or construct theories (Type C). Bullying increases the risk of suicide is a Type E theory; perceived burdensomeness increases the risk of suicide is a Type C theory. Type E theories can be proved and refuted. Type C theories cannot be proved directly. Levy noted that we can neither prove nor disprove the existence of the unconscious. Type C theories can be evaluated only on their usefulness. For example, is Einstein’s theory of gravity (a construct) more useful than Newton’s theory of gravity. Because Type C theories cannot be disproved, people mistake them as truths, and the theories survive longer than they should after they have outlived their usefulness.

Error 3: Physical and Psychological Events occur Simultaneously

            Psychological events include perceptual, experiential, cognitive and mental variables; physical variables are biochemical, physiological, anatomical and neurological. What is the relationship between these two sets of variables. Do physical events cause psychological variables? For cause-and-effect

(i)                  Event A must occur before event B

(ii)                When event A changes, event B changes accordingly.
               
Physical events can occur without psychological consequences. (For example, physical events occur after death!) On the other hand, psychological events cannot occur without physical events. When we observe a patient, the physical and psychological events occur simultaneously. Therefore, Levy argues, it makes no sense to say that a patient’s problem is physical or mental. It is always both. Furthermore, if the physical and mental variables are measured at the same time, it is not possible to say that one variable caused the other.

            In most research on suicide, researchers identify physiological (and psychological) correlates of suicidal behavior. Since these are occurring simultaneously, criterion (i) is not met. In most suicide research, event A is rarely measured before event B (suicidal behavior). Furthermore, when longitudinal studies are conducted, event A is often a construct. For example, Lester (1991) studied the gifted children followed up in the Terman study at Stanford University. Lester found that the parents’ judgment that their child at age 10 had a strong desire to excel predicted suicide later in life rather than earlier in life. The parents’ judgment cannot be said to have caused their child’s later suicidal behavior. Furthermore, the desire to excel is a construct, and so we cannot conclude that this desire caused the later suicidal behavior. On the other hand, loss of the father by death or divorce and a longer pregnancy predicted suicide at an earlier age. Here were an actual event at time 1 and a behavior at time 2. This meets criterion (i) and avoids reification (Error 2).

Error 4: The Nominal Fallacy

            This fallacy involves naming a phenomenon and then thinking that we have explained it. Levy gives an example. Why does she have difficulty falling asleep? Because she has insomnia. This is a tautology. An example of this is one of the earlier explanations for the sex difference in suicide rates. Why do men die by suicide more than women while women attempt suicide more than men do? Because men have a stronger suicidal intent. This is a tautology and not an explanation.

Error 5: Dichotomous versus Continuous Variables

            Variables can be divided into two mutually exclusive categories or they can be continuous. A person who engages in a suicidal act may either survive or die - a dichotomy. But normal-abnormal or conscious-unconscious, for example, are continua. Errors occur in theorizing when scholars dichotomize variables that are continuous. This arose, for example, in discussions as to whether suicide bombers are suicides or not. Some argue that they are not suicides (e.g., Abdel-Khalek, 2004), but the level of suicidal intent is a continuous variable, not a dichotomous variable. We should, therefore, talk of degrees of suicidal intent.

Error 6: Not Considering the Opposite

            Levy noted that in order to define a concept, we need to define its opposite. To define mental illness we need to also define mental health. Levy gave an example of considering the opposite by contrasting research into changing people’s attitudes that was stimulated by McGuire and Papageorgis (1961) who considered instead the question of how we can resist attitude change, leading researchers into the study of inoculation.

            An excellent example in suicidology is an old paper by Norman Farberow (1970) in which he addressed the problem of raising a child so as to maximize the probability that the adult would die by suicide. It provided a stimulating contrast to the typical papers on how to prevent children from becoming suicidal. More recently, after years of research on risk factors for suicide, interest has turned to protective factors. In some instance, a protective factor is simply a low score on a scale measuring a risk factor, but some constructs, such as reasons for living (Linehan et al., 1983), are unique.

Error 7: All Things are the Same; Everything is Unique

            Levy noted that, when we contrast two objects, they can have no overlap, a little overlap, a great deal of overlap, and complete overlap. When comparing two objects, no matter how much they have in common, at some point there will be a conceptual fork or bifurcation in the road after which they differ. Levy called this fork the point of critical distinction (PCD). Before this point, the two objects are similar; after this point, they are different. This results in two types of error.

Error 7a: Differences Obscured by Similarities

            Here, we let the similarities between two phenomena eclipse their differences. At a telephone hotline, the crisis counselor may decide, “Here we have another typical depressed middle-aged man (or woman),” and miss the unique features of this particular caller. The crisis counselor then inappropriately applies the usual “cookie cutter” approach.

Error 7b: Similarities Obscured by Difference

            Levy gives the example of a black client telling a white counselor that the counselor could never understand his (or her) problems. The counselor might respond, “You’re right, I can’t. But I’m a woman, and I have experienced discrimination because of my sex, and so I have had similar experiences. We are both similar and different.”

            Levy suggests always asking two questions. How are these two phenomena similar? How are these two phenomena different? For example, Lester (in press) recently asked how suicide bombing, protest self-immolation and hunger strikes are alike and how are they different.

Error 8: Confusing “Is” with “Should”

            Levy called this the naturalistic fallacy, and noted four variants: (i) if something is common, then it is good, (ii) if something is uncommon, then it is bad, (iii) if something is common, then it is bad, and (iv) if something is uncommon, then it is good. With respect to (i), Levy noted once upon a time, slavery, child labor, public torture and burning books, heretics and witches were all common. Were they good?

            Levy noted that evolutionary psychology labels behaviors that propagate the genes of the individual (or the genes of his or her family group) as “natural.” It is natural for men to seek as many young female partners as possible while women prefer monogamous relationships with rich and powerful men. Does this make the sexual double standard acceptable? High suicide rates among those unable to pass on their genes effectively helps the group. Does this make it acceptable? Examples can be found of all four variants of the naturalistic fallacy.

Error 9: Correlation does not Prove Causation

            I hope we have all learned this lesson well in our undergraduate statistics and research methods courses! However, Levy noted a variant of this in which it assumed that, because two events occurred close to each other in time, one caused the other – the contiguity-causation error. We run the risk of this by giving too much weight to the “precipitating event” when trying to understand why an individual chose to die by suicide. This error results in magical thinking and superstition, as in many athletes who wear their “lucky” clothing to improve their chances of winning.

Error 10: Failing to consider Bidirectional Causation

            As we know, a correlation between two variables A and B means that A could have caused B, B could have caused A, some third variable C could have caused both A and B, or events A and B could have a bidirectional causal loop. In the 1960s, there was a debate over whether physical punishment caused misbehaving children or whether naturally misbehaving children were so difficult to control that their frustrated parents turned to physical punishment. Rather, there could have been a causal loop (or a vicious cycle) at work. A similar bi-directionality could take place in the link between, say, drug use and depression.

Error 10: Failing to Consider Multiple Causation

            Levy called this the either/or fallacy. What is the cause of depression and suicide? Is it internalized anger, learned helplessness, or too little serotonin in the central nervous system? Levy suggested replacing “or” with “and.” Levy also noted that taking the “and” approach can lead to complex linear and nonlinear combinations of variables in our theories.

Error 11: Not all Causes are Created Equal

            It is easy to find multiple causes for a behavior if we think hard enough. Why am I a professor? The four major reasons (May, June, July and August), overcompensation for the stutter I had as a child, my exhibitionistic tendencies, etc.? Why did you, who are reading this essay, decide to study suicide? Levy suggested that each contributing cause differs in weight, degree or magnitude, and we should not neglect causes with less weight.

Error 12: Different Causes, Same Effect

            A behavior, such as depression, can be caused by many factors, such as withdrawal from drugs, vitamin deficiencies, starvation, loss, failure, loneliness, trauma, irrational thought patterns, etc.  The same applies to treatment. Depression can be ameliorated by antidepressants, cognitive therapy, supportive interpersonal relationships, etc. The error comes in assuming that similar outcomes must have similar causes.

Error 13: The Fundamental Attribution Error

            Levy defines this as our bias to attribute a behavior in an individual to internal factors and minimize external factors (the situation in which the individual finds himself or herself). If you hurt me, then you are cruel. If you fail to tip me when I serve you in a restaurant, then you are stingy. This results in our tendency to blame the victim (such as the rape victim or the battered spouse). The contrast occurs when we explain our own behavior, especially if it is behavior of which we are not proud. Then we typically hold the situation as responsible. If you do well on an exam, you take the credit for being brilliant. If you fail, you blame the examiner or some other factor in the situation that was not under your control.

            Levy suggested that the fundamental attribution error comes from our cognitive bias (in a situation, we focus on the other actors) and from our motivational bias (we endeavor to satisfy our own personal needs). Levy advises us, “Never underestimate the power of the situation” (p. 102).

Error 14: The Intervention-Causation Fallacy

            A good example of this is when you have a headache and take an aspirin. The headache goes. Did you then have an “aspirin deficiency” disease that caused the headache? Modifying an event does not, per se, prove what caused the event – the treatment-etiology fallacy. The causes of most events are multiple, and so are the ways of reversing the outcome. Individuals can become suicidal from many causes, and they can be helped to a non-suicidal state using many techniques. The method we use to help them does not necessarily indicate what caused the suicidal state (although, on some occasions, it might).

Error 15: The Consequence-Intentionality Fallacy

            This is more simply phrased as the effect doesn’t prove the intent. It may in many situations, but not always. Levy gives the example of someone who cuts their wrists severely. Can we assume that their intent was to get attention – as in Farberow and Shneidman’s (1961) classic book on attempted suicide which they called The Cry for Help? Levy suggests other possible causes, including self-punishment, sensory stimulation, confirmation of life, reification of emotion, catharsis, revenge against pain, displacement of anger, psychological control and suicide. Sylvia Plath died by suicide in 1963 in London, England, using toxic domestic gas. Did she intend to complete suicide? Alvarez (1972), one of her friend’s thought not. He argued that her behavior was a cry for help and that she expected a visitor that morning who would break in and save her. In order to avoid this error, Levy suggested that we think of other possible causes for the behavior (Error 10 above).

Error 16: Relying on Feelings

            Levy calls this the “If I feel it, it must be true” fallacy. Levy noted four possibilities here.

(i) Comfortable truths: feeling good and the event is true

(ii) Comfortable falsehoods: feeling good and the event is false

(iii) Uncomfortable truths: feeling bad and the event is true

(iv).Uncomfortable falsehoods: feeling bad and the event is false

            One of the best examples of these types of fallacies is the controversial debate over the validity of repressed memories of childhood sexual abuse. Those who believe that they have recovered such a memory or helped someone recover such a memory rely on (iii) above – if it feels bad, then it must be true. If a client of a psychoanalyst becomes uncomfortable, and even hostile, as a result of a particular interpretation made by the psychoanalyst, then this “resistance” and “defensiveness” is often used to confirm the validity of the interpretation, again an illustration of (iii).

            Levy emphasizes that one’s feelings are not an accurate or trustworthy guide to the truth.

Error 17: The Spectacular Fallacy

            This fallacy involves thinking that an extraordinary event requires an extraordinary cause. Of particular relevance to suicidal behavior, Levy notes that extraordinary human behavior (such as catatonia, hallucinations, bestiality or cannibalistic serial murder) pushes us to search for spectacular causes and to propose extraordinary theories to account for it. Levy argues that this is not a valid assumption. Extraordinary events occur sometimes by chance, as any gambler knows, or as a result of ordinary events. Psychoanalysis is based on the proposition that abnormal behavior is governed by the same principles as normal behavior. There is no qualitative difference between the two categories of behavior.

Error 18: The Pitfalls of Inductive and Deductive Reasoning

           Errors in deductive reasoning come from starting with erroneous premises and from using flawed logic. Inductive reasoning is based on data, generalizing from observations to broader principles, looking for patterns in the observations. This can lead to erroneous conclusions in several ways: (i) drawing primarily on our memory of only vivid or salient observations, (ii) ignoring statistical principles such as sample size and probability, and (iii) selectively seeking observations that are consistent with our theory and ignoring those that are inconsistent.

Error 19: Disturbing the Phenomenon by Observing It

            It is often the case that observing a phenomenon changes the phenomenon. For example, the phrasing of questions in an inventory can affect people’s responses. It has been argued that interviewing attempted suicides in the emergency room produces invalid answers to the clinician’s questions because the attempters, in all likelihood, do not wanted to be admitted to a psychiatric unit and so present themselves as hypernormal.

When interviewing survivors of those who died by suicide, the survivors may answer questions in an effort to disguise their true thoughts and feelings in order to promulgate a particular interpretation, such as avoiding admitting their own responsibility in their loved ones suicide. Lester (2013) gave a good example of this from an account by Meng (2002) of a wife, Fang, who died by suicide in China. The precipitating events for this suicide were quarrels with her in-laws and domestic violence as a victim of her husband. Her in-laws viewed Fang’s suicide as a foolish act for it cost the family a great deal in terms of cost and reputation. Fang’s parents saw Fang’s suicide as a forced decision. They blamed Fang’s in-laws, destroyed furniture in the in-laws’ house and demanded a very expensive funeral and headstone for Fang in her in-laws burial plot. The villagers gave Fang’s suicide a mystical interpretation, believing that she was taken by a ghost, which served two functions: (1) to avoid blaming Fang or her in-laws, and (2) to escape from a sense of responsibility themselves for Fang’s suicide by not intervening. The asking of questions by the investigator most likely led the interviewees to think about what the result would be for different answers that they might give and which result they preferred.

            In laboratory experiments, the researcher can sometimes use unobtrusive measures such as hidden cameras and one-way mirrors. For the study of suicide, perhaps only the study of documents, such as suicide notes and diaries, are unobtrusive ways of studying the behavior.

Error 20: Self-Fulfilling Prophecies

            Levy notes that the attitudes we have toward others can affect their behavior and certainly our judgments about those others. In victim-precipitated homicide, an individual consciously or unconsciously provokes another into killing him. In psychic homicide, an individual consciously or unconsciously encourages another to die by suicide. Some psychological research involves judges, often clinicians, making judgments about others. This is so when psychiatrists make diagnoses, but it also occurs when using judges to rate interviews or written material. Often researchers and judges are not blind to the theory and hypotheses behind the study, and this can bias the results.

Error 21: The Assimilation Bias

            Psychologists frequently categorize phenomena and behaviors, and we use schemas to do this. Our schemas are general expectations, preconceptions or theories about the phenomena we are studying. What happens when we encounter a phenomenon? If it fits into our schemas, we assimilate it.  If it does fit into our schemas, we have to accommodate, that is, shift our schemas so that now the new phenomenon fits in.

            This leads to several possible errors including, (i) noticing only that information which is consistent with our theory, (ii) selectively searching for information consistent with our theory, and (iii) distorting the information so that it fits our theory. Rosenhan’s (1973) classic study of sending normal individuals to a psychiatric inpatient unit complaining of hearing voices illustrates this bias. The eight individuals were admitted and eventually released with schizophrenia in remission after an average of 19 days (with a range of seven to fifty-two days). Everything the patients did was construed by the staff as signs of abnormality, such as taking notes and waiting for the cafeteria to open. None of the mental health personnel thought that the patients were part of a study, whereas a quarter of the patients confronted the pseudo patients and asked them why they were really in the ward.

            Levy provides a psychoanalytic joke to illustrate the assimilation bias. If a patient arrives late for a session, he is hostile; if arrives early he is anxious; and if he arrives on time, he is compulsive!!!!! Levy points out an interesting problem here. Many clinicians adopt (and perhaps believe in) a particular perspective – biomedical, psychodynamics, cognitive, behaviorist, etc. They then assimilate all information and observations into their perspective – the clinician orientation assimilation bias. Can suicidologists avoid this orientation bias?

Error 22: Confirmation Bias

            If we have a theory or a hypothesis, it may be that, when we design our research, we selectively gather information that will confirm our theory or hypothesis and we do not search for disconfirming evidence. This is called confirmation bias. Researchers show this when they select one statistical test over another because the former confirms their hypothesis better than the latter. Readers, of course, do not know how much data analysis has been tried and discarded when they read the published paper. Another form of this bias is to conduct research that only tests one’s preferred theory rather conducting research that pits your preferred theory against a rival theory.

Error 23: The Belief Perseverance Effect

            In the belief perseverance effect, researchers cling to their theory even when disconfirming evidence comes along. They discount, deny or ignore such evidence. An individual deviant can be eliminated by declaring the data from that individual to be an outlier, and sometimes the whole study can be discounted because of methodological flaws.

Error 24: The Hindsight Bias

            In a good example of this, in my early days as a research while still at graduate school, I submitted an article in which I made a clear prediction and found the opposite result. The editor, a famous psychologist, but whose name I’ll protect, rejected the paper, but gave me advice. He first suggested which journal would accept the paper, and then he told me not to predict one result and find the opposite. Predict what you found.

            How many of you, after you finished graduate school, actually choose the significance level that you will use before you run the study? How many of you had a data set, conducted a slew of analyses on the data set, and then decided what the paper would be about and what hypotheses you would test? I rest my case!

Error 25: The Insight Fallacy

            The insight fallacy is thinking that, when we understand a phenomenon, we now know how to change it. We know a great deal about the causes of suicide, but the suicide rate in the United States is steadily rising, and many suicidal individuals are in treatment but yet still die by suicide. This fallacy is also found in some therapy clients who gain insight into the causes of their problems but find it difficult to change.

Discussion

Levy’s book on critical thinking and the errors we make is a stimulating book and merits study by all researchers and theorists. I hope this essay intrigues you to read it and consider whether you have made these errors yourself.

References

 

Abdel-Khalek, A. M. (2004). Neither altruistic suicide, nor terrorism but martyrdom. Archives of Suicide Research, 8, 99-113.

Alvarez, A. (1972). The savage God. New York: Random House.

Farberow, N. L. (1970). Self-destruction and identity. Humanitas, 6, 45–68.

Farberow, N. L., & Shneidman, E. S. (1961). The cry for help. New York: McGraw-Hill.

Lester, D. (1991). Childhood predictors of later suicide. Stress Medicine, 7, 129-131.

Lester, D. (in press). Suicidal protests. Behavioral & Brain Sciences.

Levy, D. A. (2010). Tools for critical thinking: Metathoughts for psychology. Long Grove, IL: Waveland.

Linehan, M. M., Goodstein, J. L., Nielsen, S. L., & Chiles, J. A. (1983). Reasons for staying alive when you ar ethinking of killing yourself. Journal of Consulting & Clinical Psychology, 51, 276-286.

McGuire, W. J., & Papageorgis, D. (1961). The relative efficacy of various types of prior belief defense in producing immunity against persuasion. Journal of Abnormal & Social Psychology, 62, 327-337.

Meng, L. (2002). Rebellion and revenge: The meaning of suicide on women in rural China. International Journal of Social Welfare, 11, 300-309.

Robins, E., Murphy, G. E., Wilkinson, R. H., Gassner, S., & Kayes, J. (1959). Some clinical considerations in the prevention of suicide based on a study of 134 successful suicides. American Journal of Public Health, 49, 888-899.

Rosenhan, D. L. (1973). On being sane in insane places. Science, 179, 250-258.



[1] I might note in passing that I have to see a psychological autopsy study of completed suicides with an appropriate control groups of people who died of natural causes, with a standard interview protocol, and judges reading those protocols blind as to which group each individual is in (and, even better, unaware of the nature of the study and which groups are being compared).

No comments: