No one ever figured out for sure why Douglas Kelley, a successful doctor, suddenly killed himself, leaving a loving wife and three young children. (Paradoxically, in a memorial tribute to Kelley after his death, his colleagues at the University of California wrote of “his characteristically jovial humor.”) But a reasonable guess would be that as a young adult, he came to know that something about the world was terribly wrong. A sensitive person forced to face the worst of human horrors up close, he would come to see many things as far more ominous and catastrophic than would the average person. He knew firsthand how evil and scheming even educated individuals could be. Although he was outwardly successful for years after this life-changing experience, his sense of order, agency, and meaning had been shattered.
The Terman Women Who Ended It All
Our collaborator Carol Tomlinson-Keasey spearheaded a fascinating investigation of suicide, focused on eight women in the Terman sample who were definitively confirmed to have killed themselves.28 There may well have been additional suicides among the Terman women—some of the additional deaths by drug overdose were probably really suicides and not accidents. But the strongest conclusions can be made by examining those who had “suicide” listed as the cause of death on their certificates, or who had left suicide notes.
Men generally prefer guns or other violent means for taking their own lives, whereas women often turn to drugs. More than half of the confirmed female suicides among the Terman participants were by drug overdose (most often by a fatal dose of barbiturates), but there was also the occasional gunshot and carbon monoxide poisoning. Most of the suicides occurred when the women were in the prime of life—between thirty and forty-four years old. What would have led these very bright women to kill themselves? Was it an attempt to escape constant pain, anxiety, or depression by departing to a better place? Was it a very moody, impulsive temperament that led to a spur-of-the-moment decision to end it all? Was it sparked by the loss of a loved one in their lives?
In order to determine which factors were most relevant to suicide, proper comparison groups were needed. One was a matched group of women from the Terman sample who had died at the same ages (within one year) as the suicides, but from natural causes. This matched group allowed an evaluation of the typical life path of someone who died of cancer, for example, as compared to someone who swallowed a bottle of sleeping pills.
The second comparison group consisted of randomly selected female Terman subjects who lived at least until 1964, the year when the last of the scrutinized suicides occurred. This group allowed investigation into whether there might be some factors that distinguished women who died prematurely—whether by suicide or by illness—from women who survived into later life.
When we do exploratory research of this type, we want to find the relevant predictors, but we don’t want to capitalize on chance and find what was just a random quirk, never to be found again. The problem is, the more comparisons you make, the more likely you are to find some differences that came about only by chance and aren’t meaningful. That is, if researchers compare thousands of variables across the three groups of Terman participants—suicides, early natural deaths, and those who lived into later life—they would find some variables that differ substantially between the groups just by chance. To avoid this, the solution is to focus on aspects and characteristics, dealt with in other studies, that are either suspected or known to be associated with suicide.
The first of those variables is physical health. The next is stress: for this the researchers used a measure of childhood stress, a measure of adulthood stress, a coding of whether the participant had lost her father before she was twenty (a known risk factor for girls’ mental health), and a mental health index developed by Dr. Terman and his coinvestigator Melita Oden. Then came a combined measure of temperament: in 1940, were the women reporting that they were moody, unhappy, impulsive, emotional, lacking self-confidence, or feeling inferior? Finally, the researchers applied scores for what the suicidologist Edwin Shneidman called signatures of suicide—relevant elements of mental health that increase the risk of suicide. These include depression, substance abuse, anxiety, instability, and past attempted suicide.
These potential suicide predictors were then used in a statistical procedure called discriminant function analysis. The Internal Revenue Service uses discriminant function analysis to predict whether you may be a tax cheat, based on various elements of your tax return as well as previous knowledge of past cheaters. In the case of Terman suicides, this type of analysis is used to measure which elements of one’s history and pathway may predict willingness to cheat life.
The signatures of suicidal mental health did an impressive job of predicting whether someone would commit suicide, have an early natural death, or live a long life. In addition, high levels of stress and a moody, impulsive temperament contributed to the risk.
Most interesting, however, was that individual characteristics, things about the social environment, and challenges relevant to stress all contributed to suicide risk. It wasn’t just anxiety and depression, it wasn’t just alcoholism, it wasn’t just early life stress, and it wasn’t just instability in relationships and careers. All of these things mattered—a person’s overall pathway influenced the likelihood of suicide.
Many Terman subjects, for example, were anxious or faced serious challenges but lived long lives. It was those who ignored the guideposts to good health who were most at risk—they traveled down pathways that became more and more dangerous, and eventually fatal. On the other hand, those who circled back to the healthy road often went on to a long life.
Terman Men Who Shot Themselves
The late suicidologist Dr. Shneidman did not think too highly of blaming suicide on a chemical imbalance in the brain. He knew that suicide was related to deeper issues surrounding the meaning of life and death, and he knew that suicide rates varied greatly across cultures, times, and circumstances. He concluded that it couldn’t be fully understood as a brain disease, nor fully prevented with antidepressant drugs. He did not use the term catastrophizing, but he knew and agreed that the ways people thought about the world were keys to understanding suicide.
One day Dr. Shneidman called us; he started to identify himself, but we knew who he was. We knew that in addition to his other groundbreaking work, Shneidman had studied the lives of five Terman men who had shot themselves while still in their forties or fifties.29
As with the study of Terman women, this project also used two control groups—men who died at comparable ages from illness, and men who lived long lives. In a clever research move, Dr. Shneidman had an assistant prepare detailed biographies of each of the men, from childhood through about age thirty, but with all information about the man’s subsequent longevity removed. (For example, in one prepared biography, a high school counselor had written that the young man was emotionally unstable and morally erratic, while a Terman staffer who had visited the same subject described more details and reported that he was quite likable.) Shneidman then carefully read the materials in each man’s folder, without knowing which group the man was in, that is, without seeing his eventual fate. Unbeknownst to Shneidman, five were suicides by gunshot, ten were natural deaths at a comparable age, and fifteen were men still living.
28
For Dr. Tomlinson-Keasey’s study of these eight women see C. Tomlinson-Keasey, L. W. Warren, and J. E. Elliott, “Suicide among Gifted Women: A Prospective Study,”
29
To read more about Dr. Shneidman’s investigation into the lives of these men who shot themselves, see E. Shneidman, “Perturbation and Lethality as Precursors of Suicide in a Gifted Group,”