Triumphs of Experience: The Men of the Harvard Grant Study (18 page)

HEREDITY VERSUS ENVIRONMENT

In 1977 I wrote, “If isolated trauma did not affect adult life, chronically distorted childhoods did affect adult outcome. Although the mental health of relatives was not related to subsequent psychopathology in the men, the mental health of their parents was. The Worst Outcomes were twice as likely as the Best Outcomes to have a mentally ill parent. This effect would seem to be mediated environmentally.”
14

In retrospect, though, I can see that environmentalism in the postwar social sciences was just as extreme as the pre-war hereditarianism had been. This is a good example of the way scientists change science and then are changed by science in their turn. Note too that both the turn to environmentalism and the recent reaction away from it in favor of the neurosciences are encompassed in the lifespan of the Grant Study. Longitudinal studies are inescapably subject to such developments; this is one of their infuriating complications and one of their invaluable advantages. So as the years passed, I had to reconsider some of my nature vs. nurture convictions. Once again, I turned to prospective study to provide accurate data, and to reveal and redress the cultural biases of previous investigators (including myself).

For instance, data that I gathered in 2001, when the men were turning eighty, indicated that the mental health of relatives (not only parents) was an important fork in the road leading from a warm childhood to adult mental health. The presence of mental illness in the family was a wild card with the potential to trump even the most auspicious nurturance. That pointed to hereditary, rather than environmental, influence.

Table 4.1
summarizes some of that analysis. It shows mental health plotted against three conditions: bleak childhood environment, familial mental illness, and personality rating at age twenty-one. I’ve described already how we assessed childhood environment; the personality assessment at age twenty-one was part of the original investigation of the men during their college years. The scores for the second condition, familial mental illness, were established this way. In a process that will be familiar by now, we reduced four concrete indicators of genetic vulnerability—alcoholism, depression, poor familial longevity, and, surprisingly, early death of the maternal grandfather—to 3- or 4-point scales. (It can be surprisingly difficult to assess familial longevity accurately, but we were able to do it. The men’s parents knew the dates of death of
their
parents, and the Study lasted the necessary six decades until the last parent of a Study man died in 2001. Another gift of lifetime studies. Familial longevity was estimated by summing the ages of the oldest parent or grandparent on the maternal and paternal sides.)

Because these four disparate variables correlated significantly with each other, they were summed to yield a heredity score of 0 to 12, which, as
Table 4.1
illustrates, did have some predictive power. These scores did not reflect any kind of genetic testing or even precise diagnoses. They were pragmatic and crude estimates of the presence of three broad syndromes in a family—alcoholism, depression, and short life. Longevity of the maternal grandfather was included individually because its absence had a very significant—and also a fascinating and difficult to explain—association with both neuroticism and depression; much more on this in
Chapter 10
. It’s worth noting that a high heredity score was very significantly correlated with a bleak childhood, and my work on alcoholism later demonstrated that heredity accounts for almost 100 percent of familial transmission of alcoholism (see
Chapter 9
). Holmes was blessed with a heredity score of 0, while
both
Camille and Lovelace had significant genetic loading for mental illness.

Table
4.1
Important Associations of Heredity and Childhood Environment with Adjustment to Life

Very Significant = p<.001; Significant = p<. 01; NS = Not Significant.

*
This was a sum of five variables leading to vascular and heart disease (high diastolic blood pressure, Type II diabetes, obesity, heavy smoking, and alcohol abuse). The vascular risk variables are discussed in
Chapter 7
.

The first two columns of the table illustrate the correlations of childhood environment and heredity with various maturational and situational circumstances in later years. A warm childhood predicted later social and love relationships better than hereditary factors did. But heredity was the better predictor of later health-related developments
like alcohol abuse, smoking, and, most dangerously, the vascular risk variables that I’ll discuss in
Chapter 7
. Why heredity was less associated with adjustment in old age is puzzling. One factor may have been selective attrition by early death among the alcoholics and depressives.

The third column of the table reflects another curious chapter in the history of Woods’s equivocal personality trait schema—an arresting tale of a scientific ghost and the inestimable value of long perspective. In the mid-eighties, Paul Costa and Robert McCrae, longtime senior investigators at the National Institutes of Health and the Baltimore Longitudinal Study of Aging, established an inventory of personality traits that became very well known under several names: the NEO, the Big Five, and the Five Factor model.
15
That first name is an acronym for the first three of its five traits—
Neuroticism, Extraversion, Openness, Agreeableness,
and
Conscientiousness;
the second and third are self-explanatory. Among the statistically inclined, the Big Five has emerged as a robust model for understanding personality, but clinicians, including me, have found it less useful. One of my reasons is that McCrae and Costa have used it to argue that personality does not change over time.
16
This is a point of view clearly not in accord with my Grant Study experience, and one that has been sharply contested in other quarters as well.
17

In 1998 Stephen Soldz, a psychologist expert in statistics and in the Big Five, thought to look back to Woods’s earlier twenty-six-trait classification and the way it was applied to the College men in the 1940s. As I’ve said, Woods’s methodology was not all it might have been, and the usefulness of his schema has been slight, with the one startling exception that I described in the last chapter. Nonetheless, out of the material from that time, Soldz was able to extrapolate five traits that correlated, with high agreement among seven independent raters, with the Big Five. In his hands, these five traits were highly predictive
of most of the outcomes in
Table 4.1
, a result that forced me to reconsider my distrust of the Big Five’s predictive power. There was also a high correlation between Soldz’s traits and the College men’s results on the Big Five proper, which was administered to the College men at the age of sixty-seven.
18

It was fascinating to watch a measure from the very earliest days of the Study—one that for a long time looked almost laughably useless and that redeemed itself only by the skin of its teeth with the surprising success of
Well Integrated
—receive a whole new lease on life in the hands of a truly sophisticated statistician. Woods tried something. It mostly didn’t work, and the material that he amassed wasn’t much use in his own context. But if he hadn’t tried, his material would not have been available once a larger context appeared. His reach exceeded his grasp, but that overreach is an important attribute of long-sighted and imaginative science.

In another “However,” though, recent twin studies at multiple universities, especially those by David Lykken at the University of Minnesota, have shown that there is a major genetic component to Big Five test scores—perhaps up to 50 percent.
19
This implies that many personality characteristics that people tend to attribute to family and societal influences (including ones as unlikely as spirituality) are at least in part genetic. It also means that the “Bleak Childhood Environment” in
Table 4.1
reflects contributions from hereditary factors as well as environmental ones. This is a major complication for the analysis of our data, and another issue that we will have to wait for time to sort out.

The third column in
Table 4.1
refers to the NEO-related scores. It is more speculative than the other two, but it suggests that a high score on the Big Five trait of
Extraversion
(that is, thriving on challenging environments, social interactions, and keeping busy) and a low score on the trait of
Neuroticism
(that is, anxiety, hostility, depression,
and
self-consciousness) predicts a high Decathlon score. Indeed, the association (for the statisticians, rho = .45) was as strong as the benchmark correlation between people’s height and their weight.

Some of the traits that the men had been rated for in college—
Humanistic, Vital Affect, Friendly
—had been significantly associated at that time with a soundness rating of A. Others—
Inhibited, Ideational, Self-conscious and Introspective, Lack of purpose and values
—had been significantly associated at the time with a college adjustment rating of C, or unsound.
20
In midlife, however, none of these adolescent traits were associated positively or negatively either with psychopathology or with good adjustment. In retrospect, the traits associated with poor college adjustment seem to have been part of normal adolescence (which adults do sometimes view as pathological!). Surprisingly, the one trait that proved to be strongly associated with healthy midlife adjustment was a trait rather uncharacteristic of adolescents, called
Practical, Organized.
It included the ability to delay gratification, and proved to be strongly associated with healthy midlife adjustment, but not after age seventy-five. On the other hand, we found a “sleeper” trait—one that didn’t appear to be important in my first foray when the men were forty-seven, but became notably so afterward. This was our old friend
Well Integrated,
which I defined in
Chapter 2
, and which, while not so important in midlife, significantly predicted a physically active and cognitively intact old age. The reason for this appears to be that
Well Integrated
was the one Woods variable that (like a warm childhood) independently predicted the absence of vascular risk factors like smoking, obesity, and elevated blood pressure.
Self-starting
was another trait not important early in midlife but very important at the end of life.

What does it mean that a variable is predictive at one time of life but not at another? It could mean that what once looked significant really wasn’t, or vice versa. But it could also mean that some things really
are significant at certain times, but not at others. This makes prediction a much more complicated issue, and it also points up a crucial aspect of lifetime studies. They never let us forget that something that is true at one time of life may not necessarily be true at another. The longer our view, the better a chance we have to figure out why some correlations endure and others don’t. Sometimes the missing link is in our data; sometimes it’s in our science; sometimes it’s in our intuition. But even during the (admittedly uncomfortable) periods when we know a shift has taken place, or a predictor has stopped predicting, but we don’t know why, at least we know that it’s happened. It is only by studying lives over time that we become aware of changes like this, which may be markers of important developments, such as physical maturation in the brain. The vicissitudes of associations and correlations that we can follow only in longitudinal studies remind us to keep our eyes and our minds open and to guard against premature closure.

MOTHERING VERSUS FATHERING

Here’s an example of how these vicissitudes look. When I first began to analyze the effects of childhood upon adulthood, it looked like the total childhood environment was more important than the maternal relationship per se; as a psychiatrist, I found this surprising. After eighty, however, the men’s childhood relationships with their mothers look more significant—another “sleeper” effect. What its meaning might be we don’t yet know.

The Study found some facets of adulthood in which a good relationship with one parent or another exerted the more important influence. As the men approached old age, their boyhood relationships with their mothers were associated with their effectiveness at work, but their relationships with their fathers were not. A man’s maximum
late-life
income was significantly associated with a warm relationship with his mother, as was his continuing to work until seventy. His military rank and his inclusion in
Who’s Who
were also marginally significantly associated with a warm relationship with his mother.

A warm childhood relationship with his mother was significantly associated with a man’s IQ in college, and, more important, with his mental competence at eighty. A poor relationship with his mother was very significantly, and very surprisingly, associated with dementia. For example, of the 115 men without a warm maternal relationship who survived until eighty, 39 (33 percent) were suffering from dementia by age ninety. Of the surviving men with a warm maternal relationship, only 5 (13 percent) have become demented—a Significant difference. In the Grant Study, dementia has not been significantly associated with vascular risk factors. One senior colleague of mine insists that this finding must be wrong, on the grounds that no one has noted it before. He forgets that seventy-year prospective studies are as rare as hen’s teeth. Only time—or replication—will resolve the matter.