skip navigation.



Research Overview

Psychologists have long noted the similarities between infant-parent and adult romantic relationships. For example, in both kinds of relationship, people (a) feel safe and secure when the other person is present; (b) turn to the other person during times of sickness, distress, or fear; (c) use the other person as a "secure base" from which to explore the world; and (d) speak to one another in a unique language, often called "motherese" or "baby talk."

According to adult attachment theory, infant-parent and adult romantic relationships function in similar ways because they are both shaped by what John Bowlby (1969) called the "attachment system"--a motivational system designed to keep young, vulnerable infants in close proximity to potential caregivers.

The objective of my research is to explore the role that the attachment system plays in adult romantic relationships and personality functioning. Specifically, my research has addressed some of the following questions:

  • How does the attachment system manifest itself in adult romantic relationships?
  • Do securely attached children grow up to be securely attached in their romantic relationships? If so, what processes enable them to transfer a secure pattern of relating from one context to another. If not, what processes lead to change in attachment?
  • What causes some people to be secure in their close relationships? What leads others to feel insecure? How can we assess individual differences in attachment security (also known as "attachment style")?
  • How are psychological defense mechanisms used to prevent the activation of painful attachment-related feelings? How do defensive attentional processes impact the way in which defensive people experience the world?
  • What is the effect of losing a loved one on the way the attachment system functions? What are the psychological consequences for different ways of regulating the kind of emotions that are experienced when one is faced with the loss of a loved one?
  • Why does the attachment system--a motivational system that evolved in the ecology of infancy--play a role in romantic relationships? How did it evolve? What functions does it serve?

In addressing these questions, my research has drawn upon a wide range of methodologies (e.g., naturalistic observation, longitudinal methods, social-cognitive tasks, psychophysiological recordings, behavior genetic, phylogenetic analyses). In order to convey the nature of some of this work, I briefly discuss some of my research on two of these questions below. For a more detailed overview of adult attachment theory, please read the on-line primer that I have written.

The Regulation of Attachment-Related Emotion, Cognition, and Behavior

A substantial portion of my research is focused on the defensive organization of the attachment system in adulthood. My interest in this topic developed in the late 1990's when I was conducting a naturalistic study of attachment behavior among couples separating at public airports (Fraley & Shaver, 1998). The objective of the project was twofold: (a) to document the patterns of behavior observed prior to a natural, temporary separation and (b) to determine how these behaviors were related to individual differences in the kinds of attachment representations people hold. A member of my research team approached couples in local airports and asked them to complete a questionnaire designed to assess individual differences in attachment. Once the questionnaire was completed, team members and I observed each couple's behavior unobtrusively and surreptitiously. Consistent with observations on children who had been separated from their parents (Bowlby, 1969), separating partners exhibited protest behaviors (contact seeking and searching) and indications of anxiety and sadness. Interestingly, however, these kinds of behaviors were not expressed by everyone. Some adults withdrew from their partners, seeking little contact with them and providing little comfort for them. According to the self-report data, these adults were dismissing-avoidant in their attachment orientation: They had a strong desire to be independent or self-sufficient, and reported little concern over attachment-related issues.

In an attempt to better understand the regulatory systems underlying this pattern of behavior, I turned to the literature on child attachment. Children who appear cold and aloof in the Ainsworth Strange Situation procedure are classified as "avoidant" with respect to attachment. However, research on avoidant children indicates that they are much more stressed by separations than would be expected on the basis of their "detached" behavior. For example, psychophysiological evidence indicates that avoidant infants are just as distressed as secure infants during the Strange Situation (e.g., Spangler & Grossmann, 1993).

Given this evidence, there appeared to be at least two possible explanations of the aloof behavior of dismissing adults: Perhaps dismissing adults, like avoidant children, are truly distressed by separation experiences but regulate their behavior so as to appear relatively calm. Alternatively, dismissing adults might truly be less distressed by separations, and their overt behavior might match their covert emotional experience. In order to tease these two explanations apart, I used Wegner's (1989, 1994) thought-suppression paradigm to study the effect on cognition and emotion of suppressing attachment-related anxiety. Specifically, I asked people to discuss what it would be like to have their partner abandon them. Prior to discussing this topic, I had people suppress (a) the thought of their partner abandoning them (experimental condition) or (b) a neutral topic (control condition). During the experiment I recorded people's self-reported cognitive intrusions of the unwanted thought as well as their skin conductance levels. Dismissing adults were just as distressed as anyone else in the control condition. In the experimental condition, however, they were able to suppress their cognitive and emotional responses. In other words, when given the opportunity to suppress anxiety-producing thoughts, dismissing adults were able to do so relatively effectively (Fraley & Shaver, 1997).

These findings indicate that a complex set of regulatory mechanisms underlies the behavior of avoidant adults (see Fraley, Davis, & Shaver, 1998, for further discussion). Some of my ongoing research is designed to reveal more about the mechanisms that allow avoidant adults to deactivate their attachment systems. Specifically, my colleagues and I have been examining (a) the way in which attachment-related information is organized in memory (Fraley, Garner, & Shaver, 2000), (b) the way attention is allocated in response to attachment-related cues (e.g., Fraley & Brumbaugh, 2007), (c) the role of defenses in response to traumatic events (such as bereavement [Fraley & Bonanno, 2004] and the Sept 11, 2001 attack on the World Trade Center [Fraley, Fazzari, Bonanno, & Dekel, 2006]), and (d) the extent to which defenses can operate autonomously, without explicit attempts to regulate emotion and behavior. By investigating the defensive dynamics of attachment, we will learn more about the way the attachment system functions as well as the way cognition, emotion, and defenses work more generally.

Stability and Change in Attachment Security: How Stable are Attachment Patterns across the Life Course?

Many researchers who study adult attachment assume that individual differences in attachment security are stable over time. For example, it is often assumed that secure children grow up to become secure adults. However, this assumption has been hotly contested over the years and has led to major splits in the field. This issue has become especially controversial in the last few years due to conflicting findings from recently emerging longitudinal studies that have tracked attachment security from infancy to adulthood (e.g., Waters et al., 2000; Lewis et al., 2000).

Some of my recent research has been designed to address this debate. The first step in this research program was a meta-analysis that combined data from various longitudinal studies of attachment to determine how much continuity exists in security from infancy to adulthood (Fraley, 2002). These results suggested that stability exists, but that it is fairly modest over time (r's range from .10 to .39 over time). More importantly, the data indicated that the degree of stability that does exist is independent of time. In other words, infants who were classified as secure at 1 year of age were just as likely to be classified as secure at age 2 as they were to be classified as secure at age 19 (Fraley & Brumbaugh, 2004).

This kind of finding raises important questions about the developmental mechanisms that give rise to continuity and change in security. As such, the second step in this research program has been developing formal models of these mechanisms and designing studies to examine them empirically. In recent pair of theoretical papers (Fraley & Brumbaugh, 2004; Fraley & Roberts, 2005; Fraley, Vicary, Brumbaugh, & Roisman, in press) my colleagues and I formalized several developmental models and examined their implications for the patterns of continuity that should be observed over the life span. This work led to several important findings. Most importantly, it suggested that, in order to explain the existing data on stability in attachment, it is necessary to postulate an unchanging factor--a constant--that constrains the degree to which people can change over time. In some of my on-going empirical work, my students and I are trying to determine whether that "constant" is attachment-specific (e.g., based on non-verbal representations developed early in life) or attributable to basic personality traits (e.g., neuroticism). Our work in this area seems to suggest that the Big Five personality traits do not fully explain the patterns of stability that we observe in adult attachment (Fraley, Vicary, et al., in press).

Another important finding from this theoretical work is that the modus operandi in developmental research on continuity is incapable of resolving debates about which developmental mechanisms contribute to continuity. A typical study on the stability of a psychological construct, whether it be attachment security, cognitive ability, or motivation, is based on assessing that construct on two occasions and computing a test-retest correlation to quantify the stability of individual differences. By convention, if that test-retest correlation is high, researchers conclude that the construct is stable or trait-like. Our theoretical work, however, indicates that alternative developmental models--models that are often hotly contested in the literature on personality and development--could fit any empirical test-retest correlation. As such, the kind of data that is typically published on continuity and change is incapable of being used to tease apart the relative contribution of different developmental processes. In a recent set of papers (i.e., Fraley & Brumbaugh, 2004; Fraley & Roberts, 2005; Fraley et al., in press) my colleagues and I have shown what kinds of data are needed to uncover the developmental mechanisms that give rise to continuity in different kinds of psychological constructs. It is our hope that this work will help advance contemporary debates regarding stability and change not only in attachment, but in personality more generally.

With the help of a grant from the National Science Foundation (BCS-0443783), my students and I have been investigating attachment stability and change in a series of longitudinal studies on people in dating and marital relationships. One aspect of this research that we're especially excited about is the independent assessment of attachment styles in different kinds of relationships over a period of time. For example, in one of our samples we assessed attachment styles with respect to people's mothers, fathers, romantic partners, and best friends once a day over a 30-day period or weekly over a 12-month period (see Fraley, Heffernan, Vicary, & Brumbaugh, in press; Fraley, Vicary, et al., in press). This design is a unique one in this area of research because adult attachment researchers typically assess people's "global" attachment style--the way people relate to important others in general rather than the way they relate to specific individuals in their lives. This research will allow us to answer a number of important questions for attachment theory, such as, How similar are the representations that people hold of important people in their lives? If people are relatively secure with their mothers are they more likely to be secure with their romantic partners? If security changes in one relationship domain (e.g., with a romantic partner), does security in other relational domains change too? How does the breakup of a romantic relationship impact one's attachment orientation? If attachment styles change under such circumstances, how long does it take for people to "recover" and do individual differences in recovery rate reveal anything important about the organization of the attachment system?

Mom   Dad   Parnter   Friend

Mom   Dad   Parnter   Friend

To illustrate some of the data we have collected, I have created two dynamic graphs above that show the data for two of our research participants. (These graphs reply upon the Adobe Flash player. If you do not see them, the Flash player may not be installed on your computer.) The graph on the left illustrates Person A's attachment-related anxiety and avoidance with respect to four relationship domains: mom, dad, partner, and best friend. Because the ratings are made once a day over a 30-day period, the graph portrays the trajectory of a person's attachment style in that two-dimensional space over the 30-day period. For this particular individual, there is a bit of change over time. Most notably, however, the person has very different ways of relating to his best friend (with whom he is relatively secure--low in anxiety and avoidance) than he does with his parents. For example, the blue points, which represent his attachment to his mother, tend toward the extreme opposite end of the two-dimensional space. His attachment to his father is similar, but more extremely insecure and more stable.

The data portrayed on the right come from a research participant who experienced a breakup in her relationship around Day 21. It is noteworthy that this person's romantic attachment style shifts dramatically away from the secure region of the space and toward the moderately anxious and highly avoidant extreme of the space. It is also noteworthy that her attachment toward other important people (which is highly secure) does not change as a consequence of the breakup--the effect of the breakup is domain specific.

Psychometric and Measurement Issues

In modern psychology there is a growing chasm between the needs and concerns of applied researchers and the methods and tools developed by quantitative psychologists. One of my goals is to help bridge this gap by integrating contemporary developments in quantitative methodology with classical issues in the fields of social, developmental, and cognitive psychology. To date, my quantitative and methodological work has focused on four topics: (a) dynamic modeling of developmental and social processes, (b) taxometrics, (c) item response theory, and (d) Internet research methods.

A long-standing debate in the field of adult attachment has been whether variation in security is best represented by a categorical or a dimensional model. (This controversy is sometimes referred to as the "types versus dimensions" debate.) Unfortunately, this debate has been difficult to settle because previous attempts to do so have employed statistical methods that are fundamentally unable to answer the question (e.g., cluster analysis). My initial paper on this issue reported the first test of the categorical model of attachment using taxometric techniques developed by Paul Meehl and his colleagues (e.g., Waller & Meehl, 1997). Unlike other techniques that have been employed in the past, these taxometric procedures are designed to distinguish latent types from latent dimensions. In this paper my colleagues and I showed that a categorical model of attachment is inconsistent with the data and that a dimensional model best accounts for variation in adult attachment patterns (Fraley & Waller, 1998). Sue Spieker and I published an article in Developmental Psychology that reports a taxometric analysis of infant attachment patterns (Fraley & Spieker, 2003). Our conclusions were the same as those reached in the study of adult attachment: individual differences in attachment appear to be continuous rather than categorical. My colleague Glenn Roisman and I have also shown that individual differences in the Adult Attachment Interview (AAI) are continuous as well (Roisman, Fraley, & Belsky, 2007).

Although taxometric evidence indicates that attachment security is best measured within the context of dimensional models, it is debatable whether existing multi-item inventories possess the kind of precision necessary to accurately assess the full range of variation in attachment. In an article published in 2000, I illustrated the various psychometric problems that arise when scales do not measure all regions of a latent dimension with equal precision. By using polytomous item response models based on Item Response Theory (IRT), I illustrated some of the problems with existing attachment inventories and showed how IRT techniques can be used to construct better scales (Fraley, Waller, & Brennan, 2000). Our lab is currently designing new attachment measures that can be used to assess attachment across multiple contexts, including family, romantic, and peer relationships (see Fraley, Heffernan, et al., in press).

Another area of interest is the statistical significance testing debate. Significance tests have become the primary analytic technique in the toolbox of modern psychologists. However, there is growing evidence that the use of significance testing in psychology can lead to poor research design and an inability to detect true patterns in data. Many of these issues have been summarized for a social/personality audience in a chapter (Fraley & Marks, 2007).

I am also interested in how the Internet can be used to augment traditional methods for conducting social-personality research. I published a book in 2004 that discusses how to use Perl/CGI as a means for designing web-based research that enables random assignment, longitudinal assessments, and a variety of other techniques (Fraley, 2004). Our lab is currently using Internet-based tools in most of our research and trying to find new ways to make the most of web-based technology for studying personality and relationships (Fraley, 2007) Some of our web-based studies are available online at, along with a number of free personality/attachment apps that were designed for educational and demonstration purposes.