One of the strongest articles of faith among psychotherapists is the intuitively attractive proposition that the security of early attachments to parents has a profound influence on adult mental health. Thousands of articles, books, and conferences have probed this topic, and many therapists have made attachment theory a cornerstone of their clinical approach. Even clinicians who aren't particularly loyal to attachment theory accept the general proposition that the quality of infants' emotional experiences with their caretakers affects their vulnerability to psychological disorders as adults. However, when I examine the evidence for this belief as a research psychologist, rather than as a clinical practitioner, a different, less clear-cut picture emerges. Three major assumptions underlie attachment theory. First, variation in the caretaker's interactions with the infant creates variation in the infant's emotional bond to that person. This bond is called an attachment. Most psychologists, including this writer, regard this assumption as true and proven by evidence. Second, the consequences of the quality of the early attachment are preserved for many years and influence the older child and adult's personality and vulnerability to pathology. This assumption, as we shall see, is far from proven. The final premise involves the measurement of the infant's quality of attachment. As we shall see, many psychologists believe that an experimental procedure called the Strange Situation is a sensitive measure of the different types of infant attachments. This assumption, too, is a hypothesis still awaiting confirmation. A theory that rests on three assumptions, only one of which has consensual validity, should require closer examination before being embraced as indisputably true. But many clinicians believe that all three assumptions of attachment theory have been proven to be correct.

Attachment Theory in Perspective

Some influential ideas in the social sciences have their roots in the life experiences of the creator and his or her culture. This appears to be true of attachment theory. Let us consider the life experiences of the British psychiatrist John Bowlby, the father of attachment theory. As the fourth of six children growing up in an upper-middle-class London family, Bowlby, born in 1907, and his siblings were cared for by nurses on the top floor of the family's spacious home. He recalled seeing his mother for perhaps an hour each day after teatime, and his father, a prominent surgeon, once a week. His favorite nanny, with whom he had a close relationship, left the household when he was 4. By 7, he'd been sent to a boarding school. He later said of that experience, "I wouldn't send a dog away to boarding school at age 7."

Of course, we can't know whether the frequent separations from parents and the loss of the nanny contributed to Bowlby's strong ideas about attachment. We do know, however, that from the beginning of his career, he was unusually sensitive to the importance of a child's experience of parental love. He'd been trained in psychoanalytic theory and had a personal analyst. At age 21, he worked for a short time at a progressive school for emotionally disturbed children. Some of these children had experienced early separation from their parents or obvious neglect, and Bowlby interpreted their disturbed behavior patterns as support for his belief that a mother's love for a child was vital for healthy psychological development—as vital as good nutrition is for physical growth.

In 1946, Bowlby joined the staff of the famous Tavistock Clinic in London, where he met a social worker named James Robertson, who was angry at the hospital's policy of not allowing mothers to visit young children who'd been admitted for surgery—a hospital policy intended to prevent infections from spreading through the ward. Robertson noticed that children in the second year, but not those who were younger or older, became unusually distressed when they were alone on the ward lying in a crib, and he told Bowlby about his observations. Bowlby interpreted these children's distress as further confirmation of his ideas on attachment. However, Robertson disagreed, arguing that the children's distress could be avoided by simply having any adult present on the ward. This fact meant that the crying didn't reflect the fact that the children missed their mothers, but rather that they were frightened because they were alone in an unfamiliar place.

Several years later, Bowlby read about American psychologist Harry Harlow's research with infant monkeys. Harlow had discovered that infant rhesus monkeys given a choice between resting on a wire object that had a bottle with food or one covered with soft terrycloth and no food spent most of the time on the terrycloth surrogate. More important, they were less frightened by unfamiliar objects when they were resting on the cloth-covered surrogate. Harlow concluded that tactile contact was more critical than food in establishing an emotional bond between the infant and a surrogate, and that this bond awarded the infant a measure of security.

Bowlby visited Harlow and later invited him to conferences. In addition, Bowlby began to read the research in ethology suggesting that animals inherit a special sensitivity to a small set of stimuli called releasers that evoke particular behaviors. Borrowing this idea, he applied it to the mother–infant attachment by arguing that, in human infants, the face, hair, and hands of the mother acted as releasers for the infant's behaviors of grasping and clinging, and her voice and caresses released the infant's smiling and babbling. The infant's display of these responses day after day established an attachment bond between the infant and mother. These were brilliant inferences based on Harlow's research and the evidence from the ethologists, and, in my view, are correct. But, as I noted, we have to separate the fact that infants establish different types of attachment from the controversial belief that the psychological states associated with these attachments are preserved indefinitely.

A pivotal event in the history of attachment theory occurred in 1950, when the Canadian psychologist Mary Ainsworth joined Bowlby's staff at the Tavistock Clinic. Four years later, she went to Uganda because her husband had been sent there, and, with time to spare, she studied Ugandan mother–infant interactions in the natural setting of the home. In 1969, she published Infancy in Uganda, summarizing her observations. In this book, she noted that important, genetically based differences among infants influenced their level of distress to many experiences, not just separation. She concluded that it wasn't possible to separate the contributions to the quality of attachment of the infant's daily experiences from the contributions of his or her biologically based temperament. Unfortunately, she forgot this conclusion when she later joined the faculty of The Johns Hopkins University in Baltimore, where she conducted her famous study of attachment on 23 white, middle-class, mother–infant pairs who'd been observed at home many times in the first year and tested in the Strange Situation.

In this work, the 23 infants who'd been observed at home with the mother during the first year by Ainsworth's students were seen in the Strange Situation at 1 year of age. This procedure consists of a series of 3-minute episodes that last less than 30 minutes in all. In the first episode, the mother and infant are alone in an unfamiliar room containing some toys. Next a stranger enters the room. Then the mother leaves the room, so that the infant is alone with the stranger. A few minutes later, the mother returns to the room, and the stranger leaves. Minutes later, the mother departs again, leaving the child totally alone in the room. The stranger then returns to the room. In the final episode, the mother returns to the child. The critical information used to decide the quality of the infant's attachment is the child's behaviors during the two occasions when the mother prepares to leave the room and when she returns.

Ainsworth assumed that a child who cried a little as the parent went toward the door, but was easily mollified when she returned, picked the child up, and tried to soothe it, was a securely attached infant, called Type B. About two-thirds of the infants she studied showed this pattern, and most studies conducted since have affirmed this proportion. About 20 percent didn't cry when the mother left, and ignored the mother when she returned. Ainsworth assumed that the lack of interest in the mother's absence and return implied an insecure attachment, called Type A. Finally, about 15 percent of the children cried intensely when the mother left, and were so upset when she returned that they couldn't be soothed. Ainsworth presumed that this pattern, too, implied an insecure attachment, called Type C.

The second important fact was that the observations in the home revealed that the infants called securely attached were the most content, cried the least, and were the easiest to soothe. The students classified the mothers of these infants as extremely sensitive with their children. These data are the bases for Ainsworth's landmark book in 1978, Patterns of Attachment. Neither Ainsworth nor her students considered the possibility, which Ainsworth had noted in Uganda, that the Type B, securely attached infants were temperamentally the easiest babies to care for. That's why they didn't get extremely upset when the mother left them alone in the Strange Situation and were easily soothed when she returned. It's not surprising that the mothers of these infants would appear to the students to be sensitive in the home setting, since it's easy to be a sensitive parent if you have an easy baby.

Moreover, it's worth asking whether a 30-minute observation in an unfamiliar laboratory room could accurately reflect the consequences of many thousands of hours of interaction between the mother and infant in the home during the prior year. This assumption seems to me to be hard to swallow.

It's important to recognize that Bowlby and Ainsworth's work first gained currency at a time when important changes in American society made their ideas particularly attractive. These changes include the public's brooding on the horrors of World War II and the new economic structures that were transforming the traditional family that had dominated Western culture during the prior 200 years. Young mothers were now entering the workforce in unprecedented numbers and needed to find surrogate care for their babies, often in day care centers. This disruption in the time-honored form of infant care by mothers provoked a contagion of worry among professionals and parents. Could infants be cared for adequately if they didn't have the full-time attention of their mothers? This worry had a long tradition in Europe as well as America. In the middle of the 18th century, Jean-Jacques Rousseau wrote that when women are good mothers, their sons will be good husbands and fathers. Sigmund Freud thought that mothers' nursing and toilet-training practices affected adult personality. Siegfried Bernfeld, a leading Austrian-American expert on education and child development, declared in 1929 that the mind of the adult was directly related to the experiences of the opening months of life.

Thus, while Bowlby's ideas weren't novel, his theory supported the many citizens who were bothered by the changes in rearing practices because they believed that only the biological mother could provide the love every infant needed. Mother love—warm, consistent, readily available—was presumed to be the key to mental health. I, too, held that belief in late 1975, when my colleagues and I studied the impact of day care on Boston infants during their first 29 months. But during the succeeding years, my observations and study of the literature on attachment theory have led me to question its basic assumption: namely, that the quality of attachment in infancy has a permanent effect on the child, independent of future experiences or the child's social class and culture.

The evidence indicates clearly that serious neglect or abuse of infants during the first year or two can harm the child's future psychological development. No one quarrels with that statement. But it's less obvious that, for the more than 90 percent of infants who don't suffer serious abuse, the variation in mothers' sensitivity with their infants also constrains future personality, as attachment theory would want us to believe. I want to be clear on this point. There are no studies to date that have investigated the mother-infant interaction in the first year with a large group of infants from all social classes, followed these children to age 20, and demonstrated that insensitive mothers usually produced less happy adults than sensitive mothers, when social class and culture are controlled. Thus, this aspect of attachment theory remains a hunch.

During Mao Zedong's reign from 1950s to mid-1970s, millions of Chinese infants lived in austere residential day care centers and saw their mothers on an irregular schedule. I saw some of these infants in 1973, when I visited the country. Yet there's no evidence to indicate that this practice created a large number of adults with a mental illness. Mothers from northern Germany believe infants have to learn to regulate their distress, and therefore they don't go to them every time they cry—which strikes American observers as insensitive. These infants learn to control their emotions when frustrated, and, not surprisingly, they're less likely to cry when the mothers leave them in the Strange Situation, and are classified as Type A, insecurely attached. But, as with the Chinese, there's no evidence suggesting that northern German adults are especially vulnerable to psychological problems. Toward the end of World War II, Anna Freud and her colleagues studied a small group of children who'd been imprisoned in a Nazi concentration camp since their birth. Most of these children had never had a stable relation with a mother. The psychologist Sarah Moskovitz found these adults, who were scattered around the world, with the help of Anna Freud's secretary. In 1982, she published Love Despite Hate, which summarized her interviews with them. She noted their affirmation of life. There were no suicides, and only one person had been arrested for a crime. These data are inconsistent with attachment theory. I could cite many more studies leading to the same conclusion: neither behaviors in the Strange Situation nor the parent's behaviors in the first year are powerful predictors of adult personality or mental illness among children from the same social class and ethnic group. Bowlby's ideas remain a hypothesis, not a proven fact. If they're true, we'll have to wait for new investigations to demonstrate their validity.

Reconsidering Nature

While we await more evidence of the validity of attachment theory, it may help to consider the role of infant temperaments. I've spent the past 35 years researching the impact of two specific temperaments on development, and it'll come as no surprise that I claim their relevance here. Temperament refers to an inborn predisposition to experience certain feelings and display particular behaviors during the early years. My colleagues and I have found that the temperamental biases of infants are the first conditions contributing to later variation in mood and behavior. They don't determine a particular personality, but they do limit the traits that a person can acquire. The evidence suggests that a person's observable behaviors as seen by others—what Carl Jung called their "persona"—change over time. However, their private feeling tone, colored by their temperament, seems to be preserved for a longer period.

Although we're still at an early stage of identifying the potentially large number of temperamental biases, research has established that a vulnerability to distress and vigilance to unfamiliar or unexpected events, as opposed to minimal vulnerability to distress, represent two important temperaments. My colleagues and I have followed several hundred middle-class, Caucasian children from their initial assessment at 16 weeks to their 18th birthday. At 4 months, the babies were exposed to a series of harmless and unfamiliar sights, sounds, and smells. About 20 percent of the infants arched their backs, flailed their limbs, and cried in response to these unexpected events. We called these infants "high-reactive." By contrast, in response to the same stimuli, 40 percent of the infants remained relatively unruffled. We called those infants "low-reactive." I believe that these low reactives would become Ainsworth's Type B infants.

We observed these children at ages 1, 2, 4, 7, 11, 15, and 18, and found that the infants' tendency either to become easily distressed or to remain placid in the face of unexpected change endured for many, but not all, children. As teens, more high-reactive children reported a number of unrealistic anxieties over confronting crowds, strangers, unfamiliar cities, riding a subway, and the future. The low-reactive children had few such worries, and were likely to approach new situations and people with gusto. Relatively speaking, life was easier for them.

Readers might argue that these distinctive personality profiles resulted from differences in the infants' attachment to their parents. However, that's unlikely, because Doreen Arcus, a professor at the University of Massachusetts, visited the homes of high- and low-reactive infants several times in their first year, and didn't notice important differences in the mothers' sensitivity or the infants' emotional bond to the mother. In addition, we gathered physiological measures suggesting that the high-reactive children were born with a neurobiology that rendered their amygdalae more excitable than the amygdalae of the low-reactive children. Apparently, the two temperamental groups inherited different neurobiological profiles.

In 2006, my colleague Carl Schwartz of Massachusetts General Hospital, using MRI scans, began evaluating the anatomy and brain function of 135 of these children when they became 18. He found that the high-reactive children had a thicker cortex in a small area in the ventromedial cortex of the right hemisphere. This area sends impulses to the amygdala, as well as the hypothalamus, periaqueductal gray (also known as the central gray), and autonomic nervous system. Activity in these target sites explains why the older, high-reactive children continued to be vigilant to novelty and have a highly-reactive sympathetic nervous system. It also accounts for why they arched their backs so often at 4 months of age.

As these young adults lay in an MRI scanner, they were shown a series of familiar and unfamiliar faces and objects. Schwartz's preliminary, as yet unpublished, observations reveal that the high-reactive children showed greater activity to the unfamiliar stimuli than the low-reactive children in the right amygdala and right ventromedial cortex. The high-reactive 18-year-olds who showed this brain profile most clearly were the ones who'd displayed the most distress at 4 months.

I never expected to live long enough to see any neuroanatomical or neurobiological evidence to support my early hunch that the high-reactive children were born with an excitable amygdala. Schwartz's findings don't imply that infants inherit a fixed propensity for being nervous or relaxed, introverted or extroverted. They only suggest that some temperamental biases can persist, and exert an influence on the moods of some young adults. Thus, when we consider the many influences on human development, temperament must be in the mix. But a temperamental bias doesn't determine any one personality trait or symptom of mental illness, any more than the experiences of the first year of life or the legacy of a specific gene do. The power of the temperamental bias is to make it more difficult to acquire certain traits. Let me explain. The probability that a high reactive infant will become an 18-year-old who's shy, timid, and socially anxious is about 20 percent. But the probability that this infant won't grow up to be an outgoing, bold, and exuberant adult is 90 percent. Thus, the temperamental bias predicts with great accuracy what the child won't become.

Recall that this conclusion resembles the one that Mary Ainsworth arrived at in her book Infancy in Uganda. She wrote, "We must concede that there are genetically-based individual differences between babies. . . . It is quite impossible to differentiate genetic, prenatal, and perinatal influences from environmental influences." she described one infant she'd classified as "non-attached" because the girl didn't cry or attempt to follow the mother when she left the room. At the same time, this child struck Ainsworth as happy and comfortable. I suspect this girl belonged to the low-reactive temperamental group. Ainsworth also described a small group of infants classified as "insecurely attached," because they cried frequently (she'd later call this attachment Type C, insecure-resistant). But always a careful observer, she noted that two of the fussy infants were chronically malnourished—which may well have explained the constant crying. Further, she observed that many of the frequently fussy babies cried not only when their mother left their side, but also when they were held by the mother. In a telling sentence toward the end of the book, she wrote: "Therefore, the warmth of the mother and her observed affectionate contact behavior do not explain the differences between these groups." There was something else going on, and Ainsworth postulated that the "something" might be genetically based, individual differences between babies.

The Course of Human Development

A serious limitation of attachment theory is its failure to recognize the profound influences of social class, gender, ethnicity, and culture on personality development. These factors, independent of a mother's sensitivity, can be as significant as the quality of the early attachment. By the age of 4 or 5, every child perceives that he or she shares more features with some adults, especially those in the family, than with most others, based on hair style, facial attributes, skin color, and other characteristics. This perception of observable facts is accompanied by the irrational inference that perhaps the child also shares features that he or she doesn't see. For example, a 6-year-old girl with an attractive, competent, well-liked mother is tempted to conclude that she, too, possesses these traits. If the child feels moments of pride because of this inference, we say that she's identified with her mother. However, if the mother is plain, overweight, illiterate, and drinks too much, her daughter will mistakenly assume that perhaps these less desirable traits comprise parts of her personality, and she'll feel ashamed or anxious.

Consider an example of the power of identification. I recall talking with a 35-year-old Polish journalist in 2002 who grew up in anti-Semitic Poland believing that both of her parents were Roman Catholic. But when this young woman turned 20, her mother told her that she (the mother) had been Jewish, and had converted after marrying. In response, the young woman fell into a deep depression, which lasted several years. Her ethnic and religious identifications had been transformed from favored categories to those that were despised in her society. In an instant, she'd become The Hated Other. This change in belief about the self, based on discovering previously unknown information about one's family, can alter the sense of personal identity and precipitate a depression—even in an adult who's been a securely attached infant.

Research has demonstrated that social and economic factors have a powerful influence on development. The strongest predictor of adult depression or anxiety in many cultures is growing up in a disadvantaged social class. For example, Mississippi has a larger proportion of minority residents living in poverty than North Dakota, and the incidence of depression in Mississippi is three times that of North Dakota, according to the Centers for Disease Control. In fact, if two groups of psychologists were asked to predict the personality traits and incidence of pathology in 5,000 randomly selected 30-year-olds, and the first group knew only the social class in which the child had been reared, while the other group knew only the mother's sensitivity and the nature of the child's attachment during the first two years of life, the first group would make far more accurate predictions about personality and mood disorders.

Why does a disadvantaged social class position predict mood disorders, criminal careers, or addictions? One reason is that children identify with their class, which in the United States and Europe is defined by type of work, education, and income. Children belonging to less-advantaged class categories feel less potent, less virtuous, and possess a weaker sense of agency because of their identification. These traits are reflected during childhood—long before a depression or anxiety disorder develops—in lower school motivation, poorer grades, hostility to more advantaged peers, and less willingness to persist with difficult challenges.

Some influences on development that are important are less obvious, especially the person's historical era and culture. Children born in colonial America in the early 18th century were subject to harsh discipline during their socialization. But because most children experienced this form of discipline, they didn't grow up believing that their parents hated or rejected them, but became well-adjusted adults in their particular cultural setting. Shy, timid children are at risk for social anxiety in contemporary America because our culture values the outgoing, popular child who's exuberant. But this risk is far lower for children growing up in China and South Korea, where shyness and reticence are regarded as acceptable, adaptive, even praiseworthy traits.

The Researcher's Role

It's not really surprising that researchers and clinicians often speak different languages—they live in different professional worlds. As a scientist, I don't have to face the task every day of helping troubled people handle whatever challenges life has sent their way and deal with the dilemmas that have them stumped. Researchers demand a degree of empirical correspondence between ideas and data that isn't possible, or always desirable, for the mental health practitioner dealing with the unique, case-by-case demands of their practices. Of course, sometimes research can guide a clinical decision in a particular case, but more often, clinicians are faced with the idiosyncratic circumstances of a client's situation and find it difficult to always rely on a set of broad principles that tell them what to do or say. Nevertheless, psychological research can play an important role in helping clinicians do their job.

By remaining skeptical of oversimplified explanations of the human psyche and reductive answers to complex questions, psychological research forces clinicians to ask difficult questions and not pretend that they know more than they do. In that way, research serves as a corrective against the pervasive human temptation to construct a narrative that matches our preconceptions and unexamined biases, walled off from the messy reality in which all of us—clients, therapists, and researchers alike—have to live.



Adler, Nancy E., Thomas Boyce, Margaret A. Chesney, et al. "Socioeconomic Status and Health: The Challenge of the Gradient." American Psychologist vol. 49, no. 1 (January 1994): 15-24.

Beard, Courtney, Ethan Moitra, Risa B. Weisberg, et al. "Characteristics and Predictors of Social Phobia Course in a Longitudinal Study of Primary–Care Patients." Depression and Anxiety vol. 27 (2010): 839-45.

Centers for Disease Control and Prevention. "Current Depression Among Adults–United States, 2006 and 2008. Weekly Report vol. 59 ( 2010): 1229-35.

Johnson, Jeffrey G., Patricia Cohen, Bruce Dohrenwend, et al. "A Long-itudinal Investigation of Social Causation Between Socio-Economic Status and Psychiatric Disorders." Journal of Abnormal Psychology vol. 108 (1999): 490-99.

Kagan, Jerome, and Nancy Snidman. The Long Shadow of Temperament. Cambridge, Mass.: Belknap Press of Harvard University Press, 2004.

Sandmaier, Marian. "Who Do You Think You Are?" Psychotherapy Networker vol. 33, no. 3 (May 2009): 20-27, 56-57.

Schreier, Sina-Simone, Nina Heinrichs, Ronald M. Rapee, et al. "Social Anxiety and Social Norms in Individualistic and Collectivistic Countries." Depression and Anxiety vol. 27 (2010): 1128-34.

Schwartz, Carl E., et al. "Structural Differences in Adult Orbital and Ventromedial Prefrontal Cortex Predicted by Infant Temperament at 4 Months of Age." Archives of General Psychiatry vol. 67 (2010): 1-7.

Weich, Scott, Glyn Lewis, and Stephen P. Jenkins. "Income Inequality and the Prevalence of Common Mental Disorders in Britain." The British Journal of Psychiatry vol. 178 (2001): 222-227.

Photo © Karin Dreyer / Blend Images / Corbis

Jerome Kagan

Jerome Kagan, PhD, has conducted groundbreaking research on inborn temperament, personality, and the interaction of biology and psychology. His books include Temperamental Thread: How Genes, Culture, Time and Luck Make Us Who We Are, Galen's Prophecy, and An Argument for Mind.


Science & Research Families

Earn CE Credits

Just for reading the Networker!