Depression: Attachment Styles, View of Self and Negative Affect


We investigated the relationship between attachment styles and negative affect using Bartholomew and Horowitz's (1991) model of attachment. Attachment styles with a negative self view (i.e, preoccupied and fearful) were expected to be associated with more distress, especially the fearful style which involves negative views of both self and others. Measures of attachment, depression, depression proneness, and social anxiety were administered to 293 undergraduates. As predicted, participants with "negative self" attachment styles reported more symptoms of depression, proneness to depression, and social anxiety, but, contrary to prediction, those with a fearful style did not report more symptoms of depression and anxiety than those with a preoccupied style. Results suggest that the negative view of self significantly predicts depression and anxiety. Preoccupied and fearful attachment styles may best be described as predicting general negative affectivity. Implications for counseling are discussed.


Attachment theory and concepts were originally developed to address the relationship between infants and their primary caregivers (Bretherton, 1991). However, in the 1980s attachment research shifted from a primary focus on the adult-infant relationship to the application of attachment concepts to adult-adult relationships. Early attachment experiences with caregivers are believed to form prototypes or internal working models that set the stage for patterns and expectations in later adult relationships (Berman & Sperling, 1994; Rothbard & Shaver, 1994). These internal working models incorporate a view of the self as loveable or not and a perspective of others regarding whether others are likely to meet an individual's needs or provide rejection. These internal working models function as templates or attachment styles, influencing people's behaviors as they interact with and develop adult relationships, thus providing continuity between child and adult relationship patterns (Rothbard & Shaver, 1994).

Hazan and Shaver (1987) proposed a three-category model of adult attachment by applying Ainsworth's three-part model of infant attachment to describe romantic relationships in adulthood. They demonstrated that secure, anxious-ambivalent, and avoidant styles of attachment were each present in adults who differed in their histories of attachment and their cognitive, internal working models. Securely attached adults were found to perceive love positively but realistically and to view themselves as easy to get to know and like. Individuals with an avoidant style of attachment were found to fear interpersonal closeness. People with an anxious-ambivalent type of attachment were more likely to describe love with concepts of jealousy and obsessiveness and often felt less confident and misunderstood in interpersonal relationships.

Subsequently, Bartholomew and Horowitz (1991) replaced the three-category model with a four-category model of attachment. In their conceptualization of adult attachment, the view of self as positive or negative was crossed with the view of other people as positive or negative. This created four subtypes of attachment and the avoidant classification of attachment was now subdivided into two types. In the four-category model, there are secure, preoccupied, fearful and dismissive types of attachment. The securely attached people see themselves favorably and believe that other people will be responsive to them. For the fearful type, both views of self and others are negative; the person doesn't feel loveable and believes others will be rejecting and untrustworthy. For the preoccupied type (the anxious-ambivalent category in Hazan and Shaver's scheme) the person holds a positive view of others but a negative self-perspective and thus may be more "preoccupied" with relationships, i.e., often obtaining a sense of self by being valued by other people. For the dismissive type, the person has a positive view of the self but a negative view of others and thus may not seek or value relationships.

Using different models of attachment, security of attachment has been studied in relation to general psychopathology and negative mood states. Thoughts and behaviors of people with a secure style of attachment have been compared to those of people with insecure styles. For example, Mikulincer and Florian (1998) examined the coping ability of adults under stressful situations and found an insecure style of attachment was associated with more negative responses to stress, while securely attached adults managed stress more effectively and positively. In a study of college students, Kemp and Neimeyer (1999) also noted that attachment styles were associated with reactions to stress. Specifically, preoccupied attachment was correlated with more reports of negative symptoms and a feeling of distress. Administering the MMPI-2 to a low-income sample of first-time mothers, Pianta, Egeland and Adam (1996), using the Adult Attachment Interview, similarly found that attachment style predicted psychiatric symptoms; women with a preoccupied style reported the most symptoms of distress.

Other investigations of attachment and psychological distress have been applied to specific mood states, especially depression. Many studies have focused on student samples using both three-category and four-category models of attachment. Using the three-category model Roberts, Gotlib, and Kassel (1996) found that higher scores on a measure of depressive symptomatology were associated with both the anxious/ambivalent and avoidant styles, but in a regression equation only the anxious ambivalent style made a specific contribution to symptoms of depression. Using the same model of attachment, Priel and Shamai (1995) found that anxious/ambivalent students had more symptoms of depression than avoidant students. Using Bartholomew and Horowitz's four-category model, Carnelley, Pietromonaco, and Jaffe (1994) found that depression in college women was associated with both preoccupied and fearful styles of attachment.

Insecure styles of attachment also have been associated with depression in adolescent and adult samples. Interviewing psychiatrically-hospitalized adolescents with the Adult Attachment Interview, Rosenstein and Horowitz (1996) found that preoccupied and dismissive styles of attachment were correlated with specific diagnoses. A diagnosis of affective disorder was associated with a preoccupied style of attachment, while a dismissing: style was associated with conduct disorder. For married women with a history of clinical depression, responses to Bartholomew and Horowitz's model revealed that a fearful, and not preoccupied, style was associated with depression (Carnelley et al, 1994). Similarly, in a study of adult couples, Whiffen, Kallos-Lilly, and MacDonald (2001) reported that depressed women were more likely to have a fearful style of attachment than a comparison sample.

These findings on the relationship between depression and attachment are consistent with Beck's (Beck, Rush, Shaw, & Emery, 1979) cognitive triad model of depression. He proposed that people get depressed because of their negative view of the self, experience and the future. Based on Beck's conceptualization, insecure attachment styles incorporating a negative view of the self (i.e., preoccupied or fearful) would be associated with more depressive symptoms than attachment styles with a positive view of the self (i.e., secure or dismissive) (Carnelley et al., 1994). However, there appear to be inconsistent findings regarding the particular style of attachment that is most often linked with symptoms of depression. Some research suggests the anxious/ambivalent or the preoccupied style) is most vulnerable (e.g., Priel & Shamai, 1995; Rosenstein & Horowitz, 1996), while other studies propose that the fearful style best predicts depressive symptoms (e.g., Carnelley et al., 1994; Whiffen et al., 2001). Carnelley, et al., (1994) found that both preoccupied and fearful styles predicted depressive symptoms with a college student sample. This situation is further complicated since these studies used different models of attachment and varied in their use of adolescent, adult, or college student samples. In Hazan and Shaver's (1987) three-category model, the avoidant type includes people with both positive and negative views of themselves. Studies based on this model, therefore, don't directly test the fearful style of attachment and their results of specific attachment styles are then more difficult to interpret. As current research emphasizes the four-category model of attachment, more research using this model is needed to clarify the relationship between particular types of attachment styles and depression.

Furthermore, it remains unclear whether the association between depression and attachment represents a finding specific to depression or a more general finding for negative mood states or distress. For example, few studies have examined the relationship between attachment style and anxiety. Priel and Shamai (1995) used the State form of the State-Trait Anxiety Inventory in their study of college students (three-category model of attachment) and found that anxious-ambivalent students indicated more anxiety than avoidant students who, in turn, reported more anxiety than secure students. Leondari and Kiosseoglou (2000) studied university students from Greece using the Beck Anxiety Inventory and the Inventory of Parent and Peer Attachment. They found that security of attachment predicted anxiety and loneliness but they did not measure specific styles of attachment. It seems likely that participants with a negative view of themselves would also be more likely to report higher levels of anxiety, as they might worry about their performance, especially in their interpersonal interactions. However, attachment research incorporating specific types of attachment styles is needed to examine the relationship between anxiety and attachment.

The present study sought to refine our understanding of the relationship between styles of attachment, and depression and anxiety. Previous research suggests that a negative view of the self is likely to be associated with symptoms of depression. We expected to replicate this. However, prior research has produced mixed findings regarding which attachment style with a negative view of the self is most likely to be associated with depression (i.e., preoccupied or fearful). Based on Beck's theory of depression and Bartholomew and Horowitz's (1991) model of attachment, our first hypothesis was that that the fearful style, which includes a negative view of the self and others, would be most likely to be associated with symptoms of depression. The fearful style of attachment represents a more extreme form of negativity that goes beyond the self, thus expanding the person's negative perspective and experience.

In addition, we examined proneness to depression to determine if the relationship between attachment styles and symptoms of depression could be extended to include a trait perspective of depression. Most prior research has focused on immediate depressive symptoms rather than long-term patterns of depression. Our second hypothesis was that people reporting negative styles of attachment would be more likely to report being vulnerable to depressive symptoms throughout their lives.

Our third hypothesis was that attachment styles with a negative view of the self, particularly the fearful style, would be associated with other negative mood states, specifically interpersonal anxiety. Since people with the fearful style of attachment were found to be socially inhibited and lacking in assertiveness skills (Bartholomew & Horowitz, 1991), they also were expected to report higher levels of interpersonal anxiety. An attachment style incorporating a negative view of the self and others may be associated with negative mood states in general, rather than specifically linked to depression.

In summary, it was hypothesized that participants with a negative view of themselves (i.e., preoccupied and fearful) would report more symptoms of depression and interpersonal anxiety as well as a greater proneness to depression than participants with a positive view of themselves (i.e., secure and dismissive). Furthermore, it was predicted that participants with a fearful style of attachment would indicate the highest levels of depression, interpersonal anxiety, and proneness to depression.




Participants were 293 undergraduates attending either a small liberal arts college for women in the southeast (n = 149) or a private coeducational liberal arts college in the northeastern United States (n = 144). Eighty-eight percent were women and 12% were men. Regarding ethnic identity, 78% were Caucasian, 12% African-American, 2% Hispanic, 3% Asian-American, and 5% other. Ninety-five percent of the participants were single, 2% divorced or separated, and 3% married. Of the participants who were single, 60% were not dating or dating casually while 40% reported they were in a serious relationship. Data were collected over a two-year period as part of two separate studies on social relationships and attachment. Therefore the number of participants varies for the various measures in this study.


The Relationship Styles Questionnaire (RSQ; Bartholomew & Horowitz, 1991). The RSQ is based on a 4-group model of attachment proposed by Bartholomew and Horowitz. It consists of four short paragraphs, each of which describes one of the four attachment styles. Participants are asked to indicate which of the four descriptions is most characteristic of them. Empirical data support the utility of the four-group model (Bartholomew & Horowitz, 1991).

Beck Depression Inventory II (BDI-II; Beck, Steer, & Brown, 1996). The BDI-II is a 21 item self report scale measuring current depression. Each item is rated on a 0-3 scale, with higher scores corresponding to more symptoms of depression. In this study, item #9 assessing suicidal ideation was omitted. This was done after a faculty reviewer at one of the institutions expressed concern that given the anonymity of the study, there would be no way to follow up with a participant who answered the item in the affirmative. Regarding reliability, the coefficient alpha for 120 college students was .93 and the test-retest correlation for 26 outpatients was .93 after a one-week interval (Beck et al., 1996). The BDI-II is positively related to the Beck Hopelessness Scale (r = .68) and to the Hamilton Psychiatric Rating Scale for Depression (r = .71), indicating adequate validity (Beck, et al., 1996).

The Depression Proneness Rating Scale (DPRS; Zemore, 1983). The DPRS is a 13-item scale designed to assess a person's tendency to become depressed. Participants read each statement and rate each depressive symptom for frequency of occurrence over the past two years. Each item is rated on a 9-point scale with higher numbers indicating greater proneness to depression. Coefficient alpha calculated from the original subject pool of 360 students, was .83. Two-week, four-week and six-week test-retest correlations were .90, .82 and .72, respectively (Zemore & Dell, 1983). As reported in Zemore (1983), students' scores on the DPRS correlated significantly with depression-proneness ratings by their parents and peers.

Interaction Anxiousness Scale (IAS; Leary, 1983). The IAS is a 15 item self report measure of the tendency to experience subjective social anxiety independent of accompanying behaviors. Items consist of 15 statements that are responded to on a 5-point scale ranging from "not at all characteristic of me" to "extremely characteristic of me." In terms of internal consistency, all items correlate at least .45 with the sum of all other items and Cronbach's alpha exceeds .87. Eight week test-retest reliability is .80 (Leary, 1991). IAS scores correlate highly with other measures of social anxiousness and shyness (Jones, Briggs, & Smith, 1986; Leary & Kowalski, 1987). In addition, IAS scores correlate highly with self-reported anxiety in real interactions (Leary, 1983; 1986). For the present study, the title of this measure was changed to the "interpersonal interaction Scale."


The study took place at two locations: a small, private liberal arts college for women in the southeast and a small, liberal arts university in the northeast. Participants at both institutions were recruited through psychology classes and were offered extra credit for taking part in the study by the individual professors teaching those classes.

Participants were tested in small groups and were administered the above measures in a fixed order as part of a larger study. The order was as follows: informed consent; demographic questionnaire measuring age, gender, relationship status and marital status, year in college and major; Relationship Styles Questionnaire; Beck Depression Inventory-II; Depression Proneness Rating Scale; and Interaction Anxiousness Scale.



Means and standard deviations were computed for each of the target measures. Since the data were collected as part of more than one study on attachment, not all participants completed each of the measures. For the BDI-II there were 238 participants with an overall mean score of 11.19 (SD = 8.40), for the DPRS there were 133 respondents with a mean of 54.45 (SD = 19.25), while the mean score on the IAS was 41.56 (SD = 11.75) for 123 participants.

Attachment responses were coded based on Bartholomew and Horowitz's (1991) model. Forty-four percent of the sample indicated the secure description best fit their behavior in close relationships, 14% the dismissive type, 15% the preoccupied type, and 27% the fearful type. Participants reporting a secure or dismissive style of attachment were identified as "positive self" (58%) and those indicating a preoccupied or fearful style were labeled as "negative self" (42%).

Using the positive v. negative "self" groups, an independent t-test compared BDI-II scores between groups to test the hypothesis that people with a negative view of themselves would report more symptoms of depression than people with a positive self view. As predicted the "negative self" styles (i.e., fearful or preoccupied)), were associated with significantly higher BDI-II scores (M = 14.04, SD = 9.17) than those with a "positive self" view (i.e. secure or dismissive) (M = 8.98, SD = 7.02), t (236) = - 4.82, p < .001, (Cohen's effect size d = .63, (Cohen, 1988).

To examine whether the fearful style of attachment was most likely to be associated with symptoms of depression, a one-way ANOVA was conducted using attachment style as the between subject's variable and BDI-II scores as the dependent variable. The mean scores for the secure, dismissive, preoccupied and fearful styles of attachment were: 8.32 (6.79), 10.91 (7.43), 15.05 (9.37), and 13.48 (9.09). The ANOVA was statistically significant, F [sub ( 237) = 8.99, p < .001 (Cohen's effect size f = .34 (Cohen, 1988) and Scheffe's post-hoc test showed that both the preoccupied (mean difference = -6.73, p < .01) and fearful (mean difference = -5.16, p < .01) attachment groups reported significantly more depressive symptoms on the BDI-II than the secure group. Contrary to prediction, these two groups were not significantly different from each other.

Depression-proneness, a trait perspective on depression, was analyzed in the same way as the BDI-II. The group with a "negative self" view reported more proneness to depression (M = 63.25, SD = 16.81) than the group with a "positive" view of the self, (M = 48.24, SD = 18.52) t (131) = - 4.78, p < .001, (effect size d = .84). A one way ANOVA with attachment style as the between subject's factor and DPRS as the dependent variable was statistically significant, F (3,132) = 8.51, p < .001 (effect size f = .45). Scheffe's post-hoc test showed that both the preoccupied (mean difference = -19.85, p < .01) and fearful (mean difference = -12.74, p < .02) attachment groups reported more proneness to depression than the secure group, but were not significantly different from each other. In addition the dismissive group had statistically lower DPRS scores than the preoccupied group of participants (mean difference = -16.63, p < .05). The means for the specific attachment types were as follows: secure 47.54 (SD = 17.83), dismissive 50.76 (SD = 21.22), preoccupied 67.39 (SD = 21.31), and fearful 60.28 (SD = 12.16).

For interpersonal anxiety, the participants with a negative view of the self (M = 47.24, SD = 12.63) produced higher scores than participants with a positive view of the self (M = 38.62, SD = 10.15) on IAS scores, t (121) = - 4.10, p < .001(effect size d = .78). A one way ANOVA with attachment style as the between subject's factor and IAS as the dependent factor was statistically significantly, F (3, 122) = 6.20, p = .001 (effect size f = .40). Scheffe's post-hoc test demonstrated that both the preoccupied (mean difference = -9.62, p < .02) and fearful (mean difference = -9.36, p < .01) attachment groups reported significantly more anxiety than the secure group and the preoccupied and fearful groups were not significantly different from each other. The means for the specific attachment types were as follows: secure 37.76 (SD = 9.37), dismissive 41.82 (SD = 12.72), preoccupied 47.39 (SD = 12.84), and fearful 47.13 (SD = 11.75).


Consistent with our hypothesis and Bartholomew and Horowitz's (1991) four-category model, we found that people with a negative view of self (i.e., fearful or preoccupied attachment style), reported more current symptoms of depression than people indicating a positive view of the self (i.e., secure or dismissive style). This finding supports Beck's theory (Beck, 1979) regarding the central role of the self in depression. It is also consistent with previous research (e.g., Carnelley, et al., 1994; Murphy & Bates, 1997) and, given the medium-to-large effect size, suggests a meaningful relationship between a negative self view and current symptoms of depression.

However, the results did not support the prediction that people with the most depressive symptomatology would be those with a fearful style, which involves negative view of both self and others. It turns out that participants with a fearful or preoccupied style (which involves a negative view of self but a positive view of others) did not differ from each other in their report of depressive symptoms. It seems that regardless of how one sees others, it is the negative view of oneself that is most consistently linked to symptoms of depression. However, the nature of the link between negative self-view and depression may differ depending on the internal working model of other. For example, preoccupied people who are depressed may show a heightened awareness and reaction to negative social cues, which then affects their view of self and may impair their ability to see themselves objectively (Lopez, 1995). In contrast, those with a fearful style may shut out constructive social feedback or shun potentially helpful relationships, which may then further impair their self-views.

Our findings are consistent with the results of Carnelley, et al. (1994) who used the same four-category model with a college student sample. However, their sample differed from the present sample. The majority of their sample (60%) reported they were currently part of a stable relationship while, in contrast, the majority of our participants indicated they were not in a serious relationship (60%). Taken together, the results of both Studies suggest that the opinion one has of oneself plays a greater rote in self reported depression than how one feels about others, regardless of whether one is in a serious relationship. For college students, serious relationships may not have continued tong enough to provide a corrective experience that may alter the association between depression and attachment style.

Attachment styles with a negative self-concept also were more likely to be associated with a long-term proneness to depression. The pattern of results demonstrated a large effect size and was similar to the pattern reported for current symptoms of depression. Students with an attachment style incorporating a negative self view (i.e., fearful and preoccupied) reported significantly greater proneness to depression than those who had a positive self view (i.e., secure or dismissive) and again, the fearful style of attachment did not differ from the preoccupied style. These findings therefore extend our view of the relationship between depression and attachment styles to incorporate both trait and state experiences of depression. Having a long-term proneness to depression may, in turn, make one more vulnerable to developing acute symptoms. Again, this is true regardless of whether one's view of others leads one to be overly dependent on others (i.e., preoccupied style) or to lack trust in and fear involvement with others (i.e., fearful style).

Finally, we extended the findings on attachment style and depression to include social anxiety. Consistent with our prediction, those whose attachment style consisted of a negative view of self-reported more subjective social anxiety with a medium-to-large effect size. Once again, the fearful style of attachment was not associated with the most interpersonal anxiety, as both fearful and preoccupied styles differed from secure styles, suggesting again that it is a poor self (rather than other) representation that is the critical factor. These findings, however, are important in extending the research on attachment to other negative moods, this time social anxiety.

There are several conclusions that can be drawn from our findings. First, it seems clear that attachment styles with a negative view of self are associated with depression and social anxiety. However, since ours was a college student sample, it is not known whether this would also be true for a clinical sample with diagnosed affective or anxiety disorders. Second, while past research has primarily studied the relationship between attachment styles and depression, we have extended the literature to include the relationship between attachment styles and both trait depression and social anxiety. Our results suggest that having an attachment style with a poor self-view is not specific to current symptoms of depression, but actually involves a cluster of symptoms and personality traits. These include: long-term vulnerability to depression, social anxiety, and anger (Mikulincer, 1998), which together might be better termed "negative affectivity" (Feeney & Noller, 1996; Watson & Clark, 1984; Watson & Pennebaker, 1989).

Developmentally, the relationships among attachment style, self view and negative affectivity are complex. It is likely that biological and temperamental factors may set the initial stage for negative states and traits. During the course of early development, attachment experiences would then further determine whether the person develops an attachment style characterized primarily by a negative self-view. Attachment style and negative affectivity would then mutually influence one another. However, what is not yet known is the direction of the relationship, i.e., whether attachment style influences the individual's experience of negative affect or whether negative affect influences the kind of attachment style one develops. Further research is needed to investigate this relationship.

Results from the present study have important implications for the counseling process. Our results indicate that the negative view of self is an appropriate target for intervention in people with depression and anxiety, as Beck would suggest. However, when the results are interpreted within an attachment framework, which takes into account view of others as well as of self, it becomes obvious that simply focusing interventions on the self may not be enough. Because fearful and preoccupied people have different self-other schemas, it follows that their interpersonal interactions may be very different. Therefore, an effective counseling approach wilt differ for each style. For example, preoccupied people, whose symptoms may stem from being other-. oriented and overly attuned to subtle and particularly negative social cues, may need help focusing on more objective self-awareness. In contrast, fearful people, whose negative view of others causes them to disengage may need help being more connected to others; (Lopez, 1995) Therapy with fearful people may need to involve particular attention to the development of a therapeutic working alliance.

These findings must also be interpreted within a developmental framework. College students are young adults, many of whom have had limited experience in adult-adult romantic relationships. As these students develop and have more extensive relationship experiences, their specific attachment styles may be modified. It would be interesting to follow students during and beyond their college years to catalogue relationship experiences and perhaps changing attachment styles that may accompany their growth and development. In contrast to the present findings, some researchers have found that the fearful style of attachment was most likely to be associated with depression among married women (Carnelley et al., 1994; Whiffen et al., 2001).

There are several factors specific to this study which may limit the conclusions. Since the majority of the participants (88%) were women, the conclusions are more applicable to women than men. Depression in women may be of a more interpersonal nature than depression in men and it may be that attachment models of depression are more applicable to women. If attachment models of depression do apply to men, it is possible that in men, the underlying relationships between self and other are different. If this is so, therapeutic interventions need to differ for the genders. For example, depression in men may be more related to the dismissive (positive self and negative other) style and it follows that counseling approaches will need to be different than for fearful or preoccupied styles. Clearly, more research needs to be done using the four-category model to study attachment and negative mood states in men. Our study is also limited by the fact that the measures are self-report. In future studies, it will be important to gather data measuring long-term depression proneness and interpersonal anxiety from other sources, for example friends and significant others. Finally, it will be important to continue this research by using the four-category model to find whether the self/other distinction applies to other negative states, such as anger and frustration. It will also be important to find out whether the negative view of self continues to be associated with general negative affectivity or only with certain specific negative mood states and traits.