Personality Characteristics Associated With ... - SAGE Journals

5 downloads 0 Views 131KB Size Report
Inventory, and measures of response bias, self-enhancement, and self-insight. ... Keywords: romantic attachment; interview versus self-report; person-.
PERSONALITY 10.1177/0146167204264291 Gjerde et al. / ATTACHMENT, AND SOCIAL PSYCHOLOGY PERSONALITY,BULLETIN AND METHOD

Personality Characteristics Associated With Romantic Attachment: A Comparison of Interview and Self-Report Methodologies Per F. Gjerde Miyoko Onishi Kevin S. Carlson University of California, Santa Cruz This study compared the personality attributes associated with self-report versus interview assessment of romantic attachment. Twenty-three-year-olds (N = 83) completed the Romantic Attachment Interview, the Experiences in Close Relationship Inventory, and measures of response bias, self-enhancement, and self-insight. Five psychologists evaluated the participants’ personality. Although both self-report and interview assessment were related to attachment-relevant personality attributes, interview assessment was slightly more likely to explain unique variance in personality, especially regarding intrapsychic attributes. Self-enhancement was negatively related to secure attachment and positively related to dismissing attachment. The opposite pattern emerged for self-insight. A subgroup of 12 vulnerable individuals who described themselves as securely attached on self-report was judged as dismissing according to interview assessment. These individuals scored low on self-insight and high on self-enhancement and psychological vulnerability. The results are discussed with reference to the relative strengths of different measures of romantic attachment and relations among dismissing attachment, self-enhancement, and narcissism.

Keywords: romantic attachment; interview versus self-report; personality; method-comparison; normalization-hypothesis; selfenhancement

Attachment theory has emerged as a major framework

for the study of personality development. The Adult Attachment Interview (AAI; George, Kaplan, & Main, 1985/1995) enabled researchers to examine adult attachment through analysis of discourse regarding childhood family experiences. Hazan and Shaver (1987) extended attachment to include romantic relationships. With few exceptions (Bartholomew & Horowitz, 1991; Crowell, Treboux, & Waters, 1999; Furman, Simon,

Shaffer, & Bouchey, 2002), studies of romantic and peer attachment have relied on self-report. Hence, two separate traditions have emerged and their comparative validity remains contested. Adherents of the AAI argue that self-report scales measure general personality constructs (e.g., Bernier & Dozier, 2002). Furthermore, Waters, Crowell, Elliot, Corcoran, and Treboux (2002) reported that interview coherence scales were unrelated to recent self-report measures and suggested that— although each method was useful—they still behaved very differently. In contrast, Shaver and Mikulincer (in press) considered the critique from AAI researchers as exaggerated and reviewed a large amount of studies in support of their position. Authors’ Note: The National Institute of Mental Health (Grant No. MH 16080) supported this research. We thank Laura Corkery for her indispensable contribution to the development of the Romantic Attachment Interview and Avril Thorne for her insightful comments on response biases in interviews. We also thank Art Aron, Kelly Brennan, Judith Crowell, Wyndol Furman, David Harrington, Cindy Hazan, Oliver John, Eva Klohnen, Beth O’Connor, and Dominique Treboux for their prototype definitions of the romantic attachment patterns, and our dedicated research assistants, Karen Badenfort, Kellyn Bartlett, Justin Behar, Stephanie Herring, Joe Li, Zoey Luxton, Molly Millwood, Tieumi Tran, Rebecca Wade, and Stephanie Wyatt for their coding of the Romantic Attachment Interview. Special thanks are due to Shelley Kenney for her vital role in developing the Romantic Attachment Qsort. Miyoko Onishi is now at The National Center for Child Health and Development, Japan (e-mail: [email protected]). Kevin S. Carlson is at National University of Singapore (e-mail: cdtcks@ nus.edu.sg). Please direct all correspondence concerning this article to Per F. Gjerde, Department of Psychology, Social Sciences II, 1156 High Street, University of California, Santa Cruz, CA 95064; e-mail: gjerde@ ucsc.edu. PSPB, Vol. 30 No. 11, November 2004 1402-1415 DOI: 10.1177/0146167204264291 © 2004 by the Society for Personality and Social Psychology, Inc.

1402

Gjerde et al. / ATTACHMENT, PERSONALITY, AND METHOD The comparative validity of self-report versus interview assessment has typically been examined by calculating the convergence between the two methods and expresses the overlap in terms of bivariate correlations (i.e., convergence correlations). Previous research has reported low to moderate convergence correlations, in part because research has frequently compared different relationship domains (i.e., family vs. romantic attachment; Crowell, Fraley, & Shaver, 1999). To obtain an unbiased estimate of method convergence, we need to compare identical relationship domains. A Brief Overview of the Current Study Our first aim was to compare self-report and interview assessment within the same relationship domain: romantic attachment. Even high correlations between measures do not guarantee that they index identical constructs (McNemar, 1969; Ozer & Reise, 1994). Hence, convergence correlations cannot adequately determine construct overlap. Estimation of the relative validity of self-report versus interview assessment requires comparisons of how well the two methods relate to theoretically meaningful external attributes. In addition to reporting convergence correlations, we therefore examined personality attributes and related them to both self-report and interview assessment of attachment patterns. Our second aim was to examine how response biases in measures of attachment may deepen our understanding of attachment dynamics. For example, Main and Goldwyn (1998) proposed that some dismissing individuals, in their self-descriptions, “normalize” their attachment-related experiences to convince themselves and others that their feelings and experiences are customary. Consequently, some dismissing individuals would emerge as secure on self-report measures and dismissing on interview measures. This study examined response biases, with special emphasis on the normalization hypothesis. Aim I: The Comparative Validity of Self-Report Versus Interview Assessment of Attachment Different assumptions underlie different measures of attachment (DeHass, Bakermans-Kranenburg, & van Ijzendorn, 1994). The AAI seeks to “surprise the unconscious” through analysis of discourse (Main, 1991). Coherence of discourse, as operationalized by Grice’s four maxims, is considered the hallmark of secure attachment (Hesse, 1999). In contrast, self-report research has focused on consciously accessible feelings and behaviors. Brennan, Clark, and Shaver (1998) argued that accurate self-report requires only a “modicum of familiarity with one’s feelings, social behavior, and beliefs about relationships and the feedback one has received from relationship partners” (p. 68). Hesse (1999) raised

1403

the possibility that individuals might have greater conscious access to current romantic attachment experiences than to early familial attachment experiences. Therefore, the lack of method convergence may partly be explained by the tendency to compare relationship domains. Hence, we compared the romantic attachment experiences assessed both by interview and by selfreport to obtain a more accurate and unbiased estimate of the relative validity of the two methods. We sought further to go beyond convergence correlations by considering attachment-related personality characteristics associated with each of the two methods. Previous studies have demonstrated relations between attachment and self-esteem (Onishi, Gjerde, & Block, 2001), interpersonal problems (Bartholomew & Horowitz, 1991), ego resiliency, depression, hostility, and anxiety (Kobak & Sceery, 1988; Onishi et al., 2001), the Five Factor Model (FFM) of personality (Shaver & Brennan, 1992), adjustment and emotion regulation (Cooper, Shaver, & Collins, 1998), thought suppression (Fraley & Shaver, 1997), and negative and positive affect (Onishi et al., 2001). Shaver and Brennan (1992) argued that attachment patterns are related to personality functioning in ways that are not redundant with common personality traits. However, these studies have relied either on self-report or on interview assessment. Hence, a systematic comparison of the two methods vis-à-vis personality within the same relationship domain has not been conducted. This study aimed to fill this key gap in the literature. Aim II: Response Bias in the Measurement of Romantic Attachment Personality psychologists have examined the influence of response biases on self-descriptions for decades (e.g., Wiggins, 1973). Response biases refer to a propensity to respond to personality items independent of content. Biases such as social desirability (i.e., tendency to attribute social desirable attributes to self) and acquiescence (i.e., tendency to agree with an item independent of its content) have received substantial attention. If response biases exist, much of the response variance is stylistic rather than substantive and this distortion may diminish judgmental accuracy (Wiggins, 1973). Examination of response biases in attachment research is particularly relevant because self-report measures of attachment have uniformly employed meaningtransparent items. Such items generally work well if “the respondent (1) can understand the question, (2) knows the answer to the question, and (3) is willing to answer the question honestly” (Rorer, 1990, p. 698). This assumption is not always satisfied. Dismissing individuals, for example, tend to violate Grice’s maxim of quality (i.e., to be truthful; Main & Goldwyn, 1998). A common

1404

PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

dismissing strategy involves restricting access to attachment information and minimizing negative influences from attachment experiences (Dozier & Kobak, 1992). Moreover, dismissing individuals tend to lack emotional self-knowledge (Brennan & Shaver, 1990), to have narrow affective and cognitive range, and to disassociate self-descriptions from their emotions (Main & Goldwyn, 1998; Onishi et al., 2001). Thus, dismissing individuals might idealize their attachment-related experiences by normalizing their behavior and feelings (Main & Goldwyn, 1998). In late adolescence and emergent adulthood, normalization may include viewing oneself as sociable, as easy to get along with, and as having many friends to “hang out with.” These friendships, however, are often superficial, with emphasis on fun, activities, and access to material objects (Main & Goldwyn, 1998). George and West (1999) reported that self-perceptions of dismissing individuals mirror self-serving biases, not a veridical model of self. Zimmermann, Maier, Winter, and Grossmann (2001) argued that insecure individuals—even when aware of their insecure attachmentrelated feelings—still idealized themselves. When asked to provide self-descriptions of their attachment patterns, Crowell, Treboux, et al. (1999) reported that more individuals were categorized as secure and fewer as dismissing. On the AAI, the results were reversed. These studies lead us to expect that some individuals will describe themselves as secure on self-report but dismissing according to interview assessment.1 In contrast, secure individuals are believed to freely explore attachment-related thoughts and feelings and to respond in a truthful and consistent manner (Main & Goldwyn, 1998). Thus, self-descriptions provided by secure individuals should, in principle, be more valid than those provided by dismissing individuals. Support for this hypothesis comes from studies indicating that individuals who score high on private self-consciousness scales have more precise self-knowledge (Siegrist, 1996). However, secure individuals also disclose relationship difficulties. This experiential openness may lead some secure individuals to appear dismissing on self-report measures. Evaluating relations between AAI classification and self-report, Hesse (1999) noted that “not only were insecure subjects likely to term themselves secure, but secure individuals were likely to consider themselves insecure” (p. 422). Preoccupied individuals are supposed to be aware of their attachment-related feelings and relationship difficulties, however overwhelmed by anxiety they otherwise might be. They are not able to communicate their problems with clarity but tend to be truthful (Main & Goldwyn, 1998). Preoccupied individuals may therefore be moderately accurate in their self-descriptions. However, some preoccupied individuals also show self-other

confusion, evaluatory oscillation, and vague and confusing speech patterns. These individuals may find it difficult to describe themselves clearly on self-report. Thus, self-report biases may confound comparisons between self-report versus interview assessment. In contrast to self-report, attachment interviews are believed to be relatively free of response biases. To our knowledge, however, only two studies have examined response biases in the AAI. Neither study (BakermansKranenburg & van Ijzendorn, 1993; Crowell et al., 1996) found evidence for social desirability. However, because both studies used the outdated Crowne and Marlowe SDS scale (1964), this issue has not been put to rest.2 In evaluating response biases, this study included newer indices of socially desirable responding as well as measures of self-enhancement and self-insight. Expectations of the Current Study With regards to Aim I, observers assessed both interpersonal and intrapsychic aspects of personality functioning. We expected interview assessment to be slightly more predictive of the participants’ personality characteristics than self-report, especially with respect to the intrapsychic characteristics of dismissing individuals. Fewer differences were expected in the interpersonal domain. Because different methods may display different strengths and weaknesses in identifying different attachment patterns, we conducted separate analyses for secure, preoccupied, and dismissing attachment. The convergence correlations were included because their magnitude sets limits for how different the personality outcomes associated with each method can be. With regard to Aim II, dismissing participants were— due to their “normalizing” response bias—anticipated to be more strongly associated with response biases than participants characterized by the remaining two attachment patterns. Normalization is conceptually related to self-enhancement, or self-idealization. Self-enhancing individuals have been found to be vulnerable and brittle (e.g., Colvin, Block, & Funder, 1995). Thus, the subgroup characterized by normalization was hypothesized to show stronger response biases, particularly high selfenhancement and low self-insight, and also to be psychologically maladjusted. Caveats The current study is part of an ongoing longitudinal project. The data were collected from 1989 to 1990, following Hazan and Shaver’s (1987) introduction of three romantic attachment patterns: secure, avoidant, and anxious/ambivalent. Bartholomew and Horowitz (1991) subsequently reformulated Hazan and Shaver’s m o d e l b y i n tr o d u c i n g fe a r fu l a ttac h m e n t. Bartholomew’s four-component model has since gained

Gjerde et al. / ATTACHMENT, PERSONALITY, AND METHOD extensive, but far from universal, acceptance. We find Bartholomew’s distinction between the dismissing and the fearful pattern worthy of further investigation. But because the data in this study were collected under the three-component model, we used that model in our analyses. This dilemma is not unprecedented in longitudinal research (e.g., Klohnen & Bera, 1998). We chose to employ the dismissing rather than the avoidant term to be consistent with our previous research (Onishi et al., 2001) as well as with Crowell, Fraley, et al.’s (1999) Current Relationship Interview (CRI; a romantic attachment interview developed and scored according to the principles governing the AAI). Our interview also measured the three romantic attachment patterns as defined by the CRI and the AAI. METHOD

Participants The participants took part in the Block Longitudinal Study (Block & Block, 1980; Gjerde, 1995). Although most lived in urban settings, they were heterogeneous with respect to social class and parents’ educational background. About two thirds were European American, one quarter African American, and one-twentieth Asian American. The participants were enrolled in the study at age 3 and have been assessed by wide-ranging batteries of personality, cognitive, and interpersonal measures at regular intervals. This study included the 83 23-year-old (45 women, 40 men) participants who (a) had been involved in at least one committed romantic relationship, (b) had completed both interview and selfreport attachment measures, and (c) for whom personality evaluations were available. Most participants had experienced more than one committed romantic relationship at the time of the interview (M = 2.4, mode = 2, range = 1 to 6), typically before graduating from high school (M = 17.8). Relationship length ranged from 3 to 84 months, with 52 reporting at least one relationship of 2 years or more. MEASURES

Interview Assessment of Romantic Attachment The Romantic Attachment Interview (RAI). The RAI (Cockery & Gjerde, 1989) is a semistructured interview developed to assess romantic attachment in young adults. The RAI parallels the AAI (George et al., 1995) in terms of structure, but its content focused on romantic relationships, including their development, reactions to separation, the nature of recurring romantic conflicts, and emotional investment. The RAI also asked participants to supply five adjectives that described their most

1405

important romantic relationship and then to recount episodes illustrating each adjective. The interview was developed to reflect the diversity of romantic relationships typical of young adults. All participants were first asked to describe their romantic relationship status and history. The 52 participants currently involved in a romantic relationship self-evaluated as “serious” described this relationship (Sequence I). The remaining 33 participants described their most important previous relationship (Sequence II). The personality characteristics of the participants interviewed according to Sequence I did not differ from those interviewed according to Sequence II; hence, subsequent analyses treated all 85 participants as a single group. (The RAI is available from the first author.) The Romantic Attachment Q-set (RAQ). The RAQ (Carlson & Gjerde, 1996) was developed to code the RAI. The RAQ, a rationally developed set of 57 items administered in a Q-set format, was based on a thorough review of the attachment literature and current attachment measures. RAQ themes include accessibility and exploration of attachment-related experiences, affect regulation, correspondence (or contradiction) between adjectives and specific memories, idealization (or derogation) of partner, comfort with closeness, dependency on partner, investment in romantic relationships, compulsive self-reliance, fear of abandonment, and beliefs about partner’s psychological availability. The 57 Q-items were sorted into a forced sevencategory distribution according to the salience of each item in the participant’s narrative about romantic attachment. Each category included eight cards, with the exception of the middle category, which included nine cards. Ratings ranged from least characteristic to most characteristic of the participant’s narrative. (The complete RAQ is available from the authors.) Using the RAQ to Code the Romantic Attachment Interview Step 1: Developing RAQ prototypes. Expert attachment researchers used the 57 RAQ items to describe prototypically secure, preoccupied, and dismissing attachment patterns. Four experts defined each prototype. In view of the high interrater agreement (αs ranged from .90 to .93), the prototype definitions were averaged, and this composite was used in subsequent analyses. The correlation of the secure prototype was –.02 with preoccupied and –.73 with dismissing; the correlation between preoccupied and dismissing prototypes was –.47. Previous studies also have reported high negative correlations between the secure and dismissing patterns, ranging from –.58 to –.71 (Feeney, Noller, & Hanrahan, 1994). Table 1 shows the three most and least characteristic RAQ items for each attachment pattern.

1406

PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

TABLE 1: Romantic Attachment Q-Set (RAQ) Items Most and Least Characteristic of the Three Attachment Prototypes Secure RAQ Attachment Prototype Most characteristic items 3. Participant freely explores thoughts and feelings during the interview (6.75) 7. Participant speaks about own relational motivations and needs in an elaborate and understandable way (6.75) 17. Participant appears comfortable with the topic throughout the interview (6.75) Least characteristic items 28. Participant seems psychologically detached from partner/ relationship (1.00) 42. Participant fears that partner may leave him/her abruptly (1.25) 9. Participant is unable to see the relationship from partner’s point of view (1.50) Preoccupied RAQ Attachment Prototype Most characteristic items 26. Participant reacts to separations from partner with unease and discomfort (7.00) 24. Participant is jealous vis-à-vis partner (6.75) 46. Participant experiences relationship break-ups as emotionally painful (6.75) Least characteristic items 28. Participant seems psychologically detached from partner/ relationship (1.00) 48. Participant does not seek comfort or support from partner (1.00) 52. Participant feels that romantic relationships are a relatively unimportant component of his/her life (1.25) Dismissing RAQ Attachment Prototype Most characteristic items 10. Participant talks about self without reference to partner/ relationship (7.00) 28. Participant seems psychologically detached from partner/ relationship (7.00) 37. Participant is made uncomfortable by partner’s demands for closeness, support, and attention (6.75) Least characteristic items 23. Participant feels that partner is reluctant to get as close as participant would like (1.00) 1. Participant speaks about partner/relationship in an emotional manner (1.25) 5. Participant freely explores thoughts and feelings during the interview (1.50)

Step 2: Coding the interviews. Ten carefully trained senior-level psychology majors used the RAQ to code the RAIs. The coders had to complete 10 training cases and obtain an average correlation for each case of at least .50 with the coding completed by the third author. The coders were monitored during the entire coding process and provided with weekly reports on their agreement with other coders (between-coder agreement) and with the average correlation among the their own RAQ

descriptions of different participants. These latter correlations, reflecting within-coder agreement, indicated the ability of coders to discern differences among the participants’ attachment patterns. Only one coder was replaced due to idiosyncratic coding. Two coders, randomly paired from the pool of 10, worked independently. The use of a large coder pool minimized the chance of idiosyncratic agreement resulting from reliance on a few coders who may have shared a common yet invalid view of attachment. When the RAQ correlation between two coders exceeded .50, the two independently obtained assessments were averaged and the composite was employed in subsequent analyses. When the two coders did not reach this criterion, a third coder was added and the assessments of the two most consensual coders were averaged. Overall, 92% of the coder pairs reached the .50 agreement criterion (average intercoder reliability was .77). In contrast, the average within-coder agreement was only .20. The difference between these two coefficients is sufficiently large to conclude that the coders achieved differential accuracy: the ability to discern differences among the participants’ attachment patterns yet also obtain high between-coder agreement (Cronbach, 1955). Step 3: Calculating RAQ prototypes scores. We used the prototype-matching approach to develop dimensional scores for both interview and self-report measures. Onishi et al. (2001) outlined this method in detail. Each RAQ prototype composite (Step 1) was correlated with each participant’s RAQ score (Step 2) to produce three continuous prototype scores for the participant: secure, preoccupied, and dismissing. These RAQ prototype scores were used in subsequent analyses. The reliability of the three prototype scores ranged from .83 to .95. Relative to men, women were judged as more secure (t = 3.36, p = .001), more preoccupied (t = 3.47, p < .001,), and less dismissing (t = 4.19, p < .0001).3 Self-Report Assessment of Romantic Attachment The Experiences in Close Relationship (ECR) inventory. The ECR inventory (Hazan & Shaver, 1990) is a rationally developed set of 48 items that describe a person’s behavior and feelings in romantic relationships. The items provide a rich sample of the core themes thought to underlie romantic attachment, including desire for proximity, fear of rejection, reaction to separation, compulsive independence, confidence in partner’s commitment, and tendency to use the partner as a secure base. Sample ECR items include the following: “I find it relatively easy to get close to my partner,” “Being separated from my partner doesn’t bother me,” and “I rarely need or ask for my partner’s help.” A 4-point scale was used (4 = strongly agree, 1 = strongly disagree). ECR prototype scores for each of the three attachment patterns were developed

Gjerde et al. / ATTACHMENT, PERSONALITY, AND METHOD using the same prototype-matching approach as described above for the RAQ. Men described themselves as slightly more dismissing than women (p < .05). The reliability of the ECR scores ranged from .77 to .89. The ECR used in this study should not be confused with the Experiences in Close Relationships scale later described by Brennan et al. (1998). Despite content overlap, these are two different scales based on different scale construction principles. (For relations between the two versions, see Onishi et al., 2001; for extensive evaluation of the construct validity of ECR scores, see Onishi et al., 2001.) Observer Evaluations of Personality Functioning The California Adult Q-Set (CAQ). The CAQ (Block, 1961/1978) consists of 100 statements about personality, cognitive, and social functioning. Five highly trained assessors used the CAQ to evaluate each participant’s personality at age 23. None of the five assessors had provided prototype definitions, conducted the RAI, or had any previous knowledge of the participants. The assessors described each participant by arranging the Q-set items into a forced nine-step distribution from least to most characteristic. The five independent CAQ descriptions were averaged for each participant; this composite was used in the subsequent analyses. Two Ph.D.-level assessors based their evaluations on in-depth interviews; the other three—all advanced graduate students— observed the participants perform many experimental tasks (e.g., psychophysiological assessment, the Stroop procedure, an illusion of control experiment, a false consensus and conformity experiment, description of self-concept, blood pressure measurement, and a defensive rigidity procedure). Each session lasted, on the average, close to 3 hours. The results did not differ significantly for interview-based assessment and observations of the experimental tasks. The CAQ evaluations provided by the assessor who conducted the RAI were not included. The average alpha reliability of the 100 CAQ items was .70. Theory-based CAQ personality scales. Most attachment research has focused on the interpersonal domains (Mikulincer & Orbach, 1995). In their study of repressive defensiveness, Mikulincer and Orbach (1995) examined the intrapsychic sphere. The distinction between interpersonal and intrapsychic spheres seems particularly important in this study given the possibility that the two methods might be differentially sensitive to interpersonal versus intrapsychic attributes. For example, some intrapsychic processes, such as repressive tendencies, might be relatively less accessible to self-report than interpersonal behavior. Interpersonal characteristics, in contrast, may be more consciously accessible and therefore more available to self-report. For example, Shaver, Belsky, and Brennan (2000) found a particularly strong

1407

overlap between AAI and self-report of romantic attachment when they assessed the degree of comfort individuals feel depending on others or having others depend on them. In this study, three interpersonal and four intrapersonal a priori scales were constructed. All seven reflect important indicators of attachment security (Bowlby, 1969). The three interpersonal scales were psychological availability (α = .92), capacity for caregiving (α = .78), and interpersonal awareness (α = .82). These scales reflect a person’s psychological availability to others, capacity to care for others, and openness to other individuals’ feelings and actions. Caregiving, in particular, is receiving increasing emphasis in attachment theory and differences in caregiving skills and motivation have been reported (Feeney & Collins, 2001). Our caregiving scale is not intended to index the caregiving behavioral system as conceptualized by George and Solomon (1999), just the caregiving component of the attachment system. The four intrapsychic scales were susceptibility to anxiety (α = .92), repressive defensiveness (α = .90), openness to experience (α = .75), and intolerance of ambiguity (α = .70). Susceptibility to anxiety and repressive defensiveness are important in understanding how persons with different attachment patterns regulate emotions (Mikulincer & Orbach, 1995). Openness to experience is associated with the concept of secure base; it denotes that attachment security enhances the propensity to explore the environment and is possibly also related to the intrapsychic openness (i.e., openness to one’s own emotions) that is supposed to defined the internal life of secure individuals. Intolerance of ambiguity relates to the emphasis of attachment theory on rigidity versus flexibility of internal working models, with tolerance reflecting greater flexibility. Secure individuals are able to take divergent or ambiguous emotional experiences and form a coherent picture of such experiences. In contrast, insecure individuals tend to segregate divergent or ambiguous emotional experiences and have difficulty incorporating such experiences into a coherent whole (Mikulincer & Orbach, 1995). The average absolute correlation (i.e., with the directional sign disregarded) among the seven scales is .47 (SD = .18). Ego resiliency. The well-validated measure of ego resiliency (Block & Block, 1980) also was included. Lack of ego resiliency denotes vulnerability and brittleness. Measuring Socially Desirable Responding, Self-Enhancement, and Self-Insight Balanced Inventory of Desirable Responding (Version 6). Two scales developed by Paulhus (1987) were included: Self-Deceptive Enhancement (SDE) and Impression Management (IMAN). The SDE indexes the tendency to respond in self-deceptive ways of which the respondent

1408

PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

may be unaware; the IMAN indexes a tendency to deliberately manage the impression being made on others in a favorable manner. Paulhus argued that both SDE and IMAN are fundamental to the understanding of selfreport measures. Self-enhancement. This study used the self-enhancement measure developed by Colvin et al. (1995). Selfenhancement refers to “unrealistic positive selfevaluations, exaggerated perceptions of control or mastery, and unrealistic optimism” (Colvin et al., 1995, p. 1152). Nine raters evaluated each of the 100 CAQ items on a scale from 1 (very unfavorable) to 9 (very favorable) in response to the question, “How favorable or unfavorable would you regard a person who possessed this trait?” The ratings were summed across the nine raters and the 12 CAQ items that received the highest favorability ratings were identified (i.e., rater-evaluated favorability). The participants also used the CAQ to describe themselves, and the 12 items identified by each participant as being most self-descriptive also were identified (i.e., participantevaluated favorability). The self-enhancement score was obtained by subtracting rater-evaluated favorability from participant-evaluated favorability (see Colvin et al., 1995, for further description of the development of this measure and descriptive statistics). Colvin et al. showed that self-enhancement, measured in this manner, had harmful ramifications for mental health and quality of social interaction. Self-insight. The CAQ self-description items were correlated, separately for each participant, with the CAQ five-examiner composite. This procedure resulted in one convergence correlation for each participant calculated across the 100 Q-items. The composite based on the five independent evaluations (described above) was used as the criterion against which the participants’ selfdescription was judged (see John & Robins, 1994, for a perceptive discussion of the “criterion problem in selfperception”). A high convergence correlation suggested that the participant possessed high self-insight (i.e., the participant’s self-description resembled the combined evaluation of the five trained examiners). A low or negative convergence correlation reflected a self-description that differed from the examiners’ view. The average selfother correlation was .46 (SD = .22), suggesting that the participants, on average, viewed themselves similar to how they were viewed by the examiners. Self-insight was associated with observer evaluations of openness to experience and self-reports of introspection, positive emotionality, and self-consciousness. This measure differed substantially from the self-enhancement measure described above. The self-enhancement measure included only the 12 items judged to be most favorable in evaluating a person. In contrast, the self-insight

measure evaluated self- versus examiner differences across all 100 CAQ items, containing negative, neutral, and positive attributes. The Q-sort is a modified ranking procedure. As such, it is particularly suited for self-other comparisons because “it provides an explicit context of comparison . . . and eliminates any potential differences in scale usage between the assessment staff and the subjects involved in the task” (John & Robins, 1994, p. 209). Separating the influence of ECR and RAQ prototype scores. Given that corresponding attachment prototype scores were correlated, regression analyses were used to determine their unique contribution to personality functioning. The SAS GLM procedure was used to generate residualized measures for both ECR and RAQ prototype scores, separately for each attachment pattern. For example, the residualized measure of the secure RAQ prototype score controls for the influence of the secure ECR prototype score. That is, the variance associated with the ECR measure is removed from the comparable RAQ measure before its association with the personality attributes is evaluated. This procedure was then repeated for the ECR measures. The method permits examination of the personality attributes uniquely associated with each of the two prototype scores (RAQ vs. ECR). We refer to these residualized scores—expressed as squared semipartial correlation—as “pure” ECR and RAQ prototype scores. It is important to keep in mind the difference between prototype scores and “pure” prototype scores. Prototype scores generate zero-order correlations; “pure” prototype scores are a residualized estimate that explains variance in a predicted variable over and above the variance explained by a second predictor variable. Partial correlations only completely reduce the influence of the variable being controlled for if this variable is reliably measured (Cliff, 1987). The variables included in this study were all highly reliable. RESULTS

Gender differences in correlational patterns were not observed, and no correlational differences due to current relationship status were discerned. Thus, only results for the full sample are reported. Method Convergence Between Corresponding Continuous Attachment Prototype Scores The convergence correlations between the RAQ and ECR prototype scores were .36 (p < .01) for secure prototype scores, .58 (p < .001) for preoccupied prototype scores, and .52 (p < .001) for dismissing prototype scores. These correlations were, on average, slightly higher than those reported by Crowell, Fraley, et al. (1999) for com-

Gjerde et al. / ATTACHMENT, PERSONALITY, AND METHOD

1409

parisons of the same relationship domain using different methods.

scores explained no unique variance for any interpersonal scales.

Method Differences in Personality Functioning: Zero-Order Correlates

Intrapsychic functioning. The RAQ pure secure prototype scores were related to all four intrapsychic scales. The ECR pure secure prototype scores, in contrast, were associated only with susceptibility to anxiety. The RAQ pure dismissing prototype scores were related to all the intrapsychic scales except susceptibility to anxiety. Neither the ECR pure dismissing prototype score nor any of the preoccupied pure prototype scores explained unique variance for any of four intrapsychic scales.

Interpersonal functioning. Table 2 shows that both RAQ and ECR secure prototype scores were positively associated with all three interpersonal scales. The RAQ dismissing prototype scores were negatively associated with these three measures of interpersonal functioning, whereas the ECR dismissing prototype scores were negatively associated with capacity for caregiving and interpersonal awareness. The two preoccupied scores were unrelated to the measure of interpersonal functioning. The differences between the RAQ and ECR zero-order correlations were not statistically significant. Intrapsychic functioning. RAQ secure prototype scores were negatively associated with susceptibility to anxiety, repressive defensiveness, and intolerance of ambiguity and positively associated with openness to experience. For secure attachment prototype scores, the difference between the zero-order RAQ and ECR correlations with repressive defensiveness was itself statistically significant (t = 2.37, p < .05). The RAQ dismissing prototype scores were positively associated with repressive defensiveness and intolerance of ambiguity and negatively associated with openness to experience. In contrast, the ECR dismissing prototype scores were not significantly associated with any of the four measures of intrapsychic dynamics. For dismissing attachment prototype scores, the difference between RAQ and ECR zero-order correlations with repressive defensiveness was itself significant (t = 2.63, p < .01). Both preoccupied prototype scores were correlated with susceptibility to anxiety (r = .30 for RAQ, r = .31 for ECR, p < .05). Ego resiliency. Table 2 also shows that both the RAQ and the ECR secure prototype scores were positively associated with ego resiliency. RAQ dismissing prototype scores were negatively related to ego resiliency. Preoccupied prototype scores were unrelated to ego resiliency. Method Differences in Personality Functioning: “Pure” Measures Interpersonal functioning. The RAQ pure secure prototype scores explained variance in all three interpersonal scales over and above the corresponding ECR secure prototype scores. The ECR pure secure prototype scores explained unique variance only for capacity for caregiving. The RAQ pure dismissing prototype scores explained unique variance for psychological availability and interpersonal awareness. The ECR pure dismissing prototype

Ego resiliency. Both the pure RAQ and ECR secure prototype scores were associated with ego resiliency. The RAQ pure dismissing prototype score also was related to ego resiliency. Relations Between and Romantic Attachment Prototype Scores and Response Bias Zero-order correlates. Table 3 shows that both RAQ and ECR prototype scores were independent of Paulhus’s (1987) SDE and IMAN response bias scales. These measures may not be sufficiently sensitive to response distortions in the attachment-related domains, in part, because these distortions might not be unambiguous and straightforward. Shaver and Brennan (1992), for example, reported that dismissing individuals were open in terms of values (especially sexual values) but not in terms of attachment-related feelings. Dismissing individuals are particularly likely to deny attachment-related negative affect (Kobak & Sceery, 1988; Shaver & Brennan, 1992). Denial of negative affect may not be sufficiently tapped by Paulhus’s two scales. Table 3 also shows that secure RAQ prototype scores were positively associated with self-insight and negatively associated with self-enhancement; in contrast, RAQ dismissing prototype scores were negatively associated with self-insight and positively associated with selfenhancement. The ECR preoccupied prototype scores were unrelated to self-insight and self-enhancement. Semipartial correlates. Both the secure and the dismissing pure RAQ prototype scores explained unique variance in self-insight and self-enhancement. The ECR pure preoccupied prototype scores were unrelated to self-insight and self-enhancement. Evaluating the Normalization Hypothesis: MethodCongruent Versus Method-Incongruent Individuals Next, we sought to identify a subgroup of individuals vulnerable to normalization. We employed the procedure used by Bartholomew and Horowitz (1991) and Mickelson, Kessler, and Shaver (1997). That is, after standardizing the scores, we derived separate RAQ and ECR categorical attachment scores by assigning participants

1410

PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

TABLE 2: Zero-Order and Squared Semipartial Correlations of RAQ and ECR Prototype Scores With California Adult Q-Sort Scales (CAQ) Attachment Pattern Secure Attachment Personality CAQ Scale Interpersonal CAQ Scales Capacity for caregiving Psychological availability Interpersonal awareness Intrapsychic CAQ Scales Susceptibility to anxiety Repressive defensiveness Openness to experience Intolerance of ambiguity Ego-resiliency

a

b

Dismissing Attachment a

b

RAQ

RAQ

ECR

ECR

RAQ

RAQ

ECR

ECR

r

sr

r

sr

r

sr

r

sr

2

2

2

2

.33** .30** .43***

.05* .05* .14***

.43*** .26* .24*

.11** .03 .01

–.29** –.27* –.41***

.03 .05* .12**

–.28** –.13 –.23*

.02 .01 .01

–.33** –.41*** .31** –.32** .42***

.05* .15*** .09** .09** .10**

–.37*** –.12 –.08 –.14 .34**

.07* .00 .00 .01 .05*

.14 .39*** –.25* .30** –.29**

.01 .12*** .06* .06* .06*

.11 .14 –.09 .19 –.16

.00 .00 .00 .00 .00

NOTE: RAQ = Romantic Attachment Q-Sort, ECR = Experiences in Close Relationship Scale. The results for the preoccupied prototype score are provided in the text. a. The variable controlled for was the corresponding ECR attachment prototype score. b. The variable controlled for was the corresponding RAQ attachment prototype score. *p < .05. **p < .01. ***p < .001.

TABLE 3: Zero-Order and Squared Semipartial Correlations of RAQ and ECR Prototype Scores With Response Biases, Self-Enhancement, and Self-Insight Attachment Pattern Secure Attachment Variable Response Bias Socially desirable responding Self-deception Impression management Self-enhancement Self-insight

a

Dismissing Attachment

RAQ

RAQ

ECR

ECR

RAQ

RAQa

ECR

r

sr

r

sr

r

sr

r

sr2

.01 .18 –.57*** .32**

.00 .03 .14** .07*

.13 .05 –.10 .21

.02 .00 .01 .01

.01 –.13 .18 –.04

.00 .01 .01 .01

2

b

2

.01 –.20 .48*** –.25*

2

.00 .02 .20** .07*

ECRb

NOTE: RAQ = Romantic Attachment Q-Sort, ECR = Experiences in Close Relationship Scale. The results for the preoccupied prototype score are provided in the text. a. The variable controlled for was the corresponding ECR attachment prototype score. b. The variable controlled for was the corresponding RAQ attachment prototype score. *p < .05. **p < .01. ***p < .001.

to one of the three attachment patterns according to their highest prototype score. Individuals who received their highest standardized prototype score on both measures on the same attachment pattern were labeled method-congruent; individuals who received their highest standardized prototype score on different attachment patterns were labeled method-incongruent. Table 4 summarizes the results. Descriptive statistics. According to their highest RAQ score, 58 participants (68.2%) were classified as secure, 8 (9.4%) as preoccupied, and 19 (22.4%) as dismissing. According to their highest ECR score, 64 participants (75.3%) were categorized as secure, 13 (15.3%) as

preoccupied, and 8 (9.4%) as dismissing. The distribution obtained for the RAQ-based category assignments is similar to the distribution Mickelson et al. (1997) reported for a large, nationally representative sample: 59% secure, 11% anxious, and 25% avoidant. In contrast, the distribution obtained for the ECR-based category assignments was skewed toward secure attachment and away from dismissing attachment. The chi-square was statistically significant (χ2 = 33.15, p < .001). Sixty-one (71.2%) of the 85 participants were method-congruent: 50 as secure, 6 as preoccupied, and 5 as dismissing. Twenty-four individuals (28.2%) were method-incongruent. Of the 24 method-incongruent

Gjerde et al. / ATTACHMENT, PERSONALITY, AND METHOD TABLE 4: Cross-Tabulation of RAQ and ECR Attachment Categories ECR (SelfReport) Category Secure Preoccupied Dismissing Column total

RAQ (Interview) Category Secure

Preoccupied

Dismissing

Row Total

50 (86.2%) 5 (8.6%) 3 (5.2%)

2 (25%) 6 (75%) 0 (0%)

12 (63.2) 2 (10.5%) 5 (26%)

64 (75.3%) 13 (15.3%) 8 (9.4%)

58 (68.2%)

8 (9.4%)

19 (22.4%)

85 (100%)

NOTE: RAQ = Romantic Attachment Q-Sort, ECR = Experiences in Close Relationship Scale. χ2 = 33.15 (p = .001). Numbers in parentheses are column percentages for each cell. Total percentages are listed in parentheses after each column and row total.

participants, 14 described themselves on the ECR as secure, 7 as preoccupied, and 3 as dismissing. According to the RAQ, 8 were evaluated as secure, 2 as preoccupied, and 14 as dismissing. The number of participants classified as dismissing differed significantly according to method (χ2 = 4.25, p < .05); 12 participants described themselves as secure but were evaluated as dismissing according to the RAQ. Significant classification incongruence was not obtained for the remaining two attachment patterns. The results should be interpreted with caution given that several cells have expected counts of less than five.4 Personality Characteristics of Method-Incongruent Versus Method-Congruent Individuals To test the normalization hypothesis, the personality characteristics of the 12 method-incongruent participants who described themselves as secure but were judged as dismissing according to interview assessment were compared to the method-congruent secure individuals. Because the size of the bipoint-serial correlations depends on the distribution of the dichotomized variable, we generated three contrast groups of methodincongruent participants to improve the robustness of these comparisons: Contrast Group I consisted of all 50 method-congruent secure participants, Contrast Group II consisted of the 12 most highly secure methodcongruent participants (both prototype scores equaled or exceeded .60), and Contrast Group III consisted of three groups of 12 randomly selected methodcongruent secure participants. Separate contrasts were conducted for each of these three randomly selected contrast groups, and the resulting three correlations obtained for each CAQ scale were averaged using r-to-zto-r transformations. Table 5 shows that the results were highly consistent across the three contrast groups. Method-incongruent participants were more likely than method-congruent secure participants to be viewed as lacking in capacity

1411

TABLE 5: Method-Incongruent Versus Method-Congruent Individuals: Differences in Interpersonal Behavior and Intrapsychic Dynamics Contrast Contrast Contrast b c d Ia II III Average rpb Interpersonal CAQ Scales Capacity for caregiving Psychological availability Interpersonal awareness Intrapsychic CAQ Scales Susceptibility to anxiety Repressive defensiveness Openness to experience Intolerance of ambiguity Ego resiliency Socially desirable responding Self-deception Impression management Self-enhancement Self-insight

rpb

rpb

rpb

–.35** .15 –.54***

–.48** .28 –.64***

–.45 .13 –.61

–.43 .19 –.60

.30* .36** –.21 .34* –.33*

.41* .52** –.38* .56** –.43*

.30 .46 –.22 .40 –.46

.34 .45 –.27 .44 –.40

–.01 –.04 .50*** –.54***

–.06 –.30 .74*** –.40*

–.00 .13 .67 –.37

–.02 –.07 .67 –.44

NOTE: We used the Fisher r-to-z-to-r transformation whenever we averaged correlation coefficients. CAQ = California Adult Q-Sort, RAQ = Romantic Attachment Q-Sort, ECR = Experiences in Close Relationship Scale. Method-congruent individuals are coded 1; methodincongruent individuals are coded 2. a. Contrast sample includes all 50 individuals who received their highest score for secure romantic attachment on both RAQ and ECR. b. Contrast sample includes the 15 individuals whose score exceeded .60 on both RAQ and ECR; these 15 can be viewed as consistently the most strongly securely attached 15 individuals in the whole sample, as measured by both interview assessment and self-report measurement. They are the most highly method-congruent individuals in the sample. c. Correlations are based on the average of three randomly drawn subsamples, each consisting of 12 individuals who received their highest score for secure romantic attachment on RAQ and ECR. Because these correlations are averaged across three samples, significance levels are not included. The average correlation is calculated using r-to-zto-r transformation. d. The average correlation for the three preceding contrasts. The average correlation is calculated using r-to-z-to-r transformation. *p < .05. **p < .01. ***p < .001.

for caregiving, interpersonal awareness, openness to experience, and ego resiliency. Method-incongruent participants were also more than method-congruent secure individuals to be susceptible to anxiety, prone to defensiveness, and intolerant of ambiguity. Methodincongruence, hypothesized to reflect “normalization,” was negatively associated with self-insight and positively associated with self-enhancement. These analyses suggest the existence of a subgroup of vulnerable participants who defensively overestimate their romantic attachment security when they describe themselves. Hence, one source of method incongruence—and possibly a source of error in self-report measures—may

1412

PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

derive from dismissing individuals who skew their selfdescriptions in the direction of security.5 DISCUSSION

Personality Functioning Associated With Romantic Attachment: Interview Versus Self-Report RAQ interview prototype scores were more likely to contribute to variance in the CAQ personality scales than the ECR self-report prototype scores. This method difference was particularly pronounced for (a) the dismissing prototype scores and (b) inferences about intrapsychic functioning, especially with regard to repressive defensiveness. Moreover, RAQ pure secure prototype scores explained unique variance in all eight personality constructs; ECR pure secure prototype scores explained unique variance only for capacity for caregiving, susceptibility to anxiety, and ego resiliency. Although the difference between secure and dismissing attachment was only statistically significant for repressive defensiveness, the overall pattern of findings is consistent with our expectation that intrapsychic processes of dismissing individuals were less accessible to self-report than to interview assessment (χ2 = 6.83, p < .01). Nonetheless, although RAQ dismissing prototype scores were more strongly related to the CAQ personality scales than ECR dismissing prototype scores, the ECR prototype scores provided useful information about personality functioning, especially in the interpersonal domain. RAQ dismissing prototype scores were also more likely to contribute to variance in self-insight and selfenhancement than the ECR prototype scores. This difference was obtained both for zero-order and semipartial correlations. RAQ and ECR preoccupied prototype scores showed positive relations only to susceptibility to anxiety. It merits attention that the overriding quality of preoccupied attachment is anxiety. This anxiety may be so overwhelmingly salient to observers that it dominates the stimulus value of the preoccupied person. Further support for this conjecture derives from the analyses conducted by Onishi et al. (2001), who correlated the three ECR prototype scores with all 100 CAQ items. In that study, 80% of the items correlated with the ECR preoccupied prototype score were related to anxiety and vulnerability. Given that many CAQ items represent anxiety and vulnerability, the salience of anxiety may have permeated the observer ratings of preoccupied individuals. Method-Incongruence Between Different Romantic Attachment Prototype Scores The convergence correlation for the dismissing prototype scores was higher (r = .52) than expected. Griffin and Bartholomew (1994) also found moderate method

convergence for the dismissing pattern (r = .40). Dismissing RAQ prototype scores were related to individual ECR items indicative of not seeking partner’s help and support, not telling partner about one’s life, not needing to know partner’s feelings, and ease of separation. Thus, these aspects of the dismissing pattern may be consciously accessible and can be self-described with considerable accuracy. Bartholomew and Horowitz (1991) argued that dismissing individuals do not constitute a homogeneous group. For example, some may be able and willing to devalue the import of attachment relationships by describing their relational counterdependence; others—here referred to as method-incongruent—may “normalize” self-descriptions in the direction of sociability to avoid the anxiety that an open admission of a dismissing pattern might provoke. This infectivity of dismissing individuals in the secure pattern may, in part, have contributed to the relatively low method convergence for secure attachment. Dismissing Attachment and Self-Enhancement: The Role of Narcissism According to interview assessment, dismissing attachment was negatively related to self-insight and positively related to self-enhancement. These results received further support in the qualities that characterized the method-incongruent individuals (i.e., securely attached according to self-descriptions; dismissingly attached according to interview assessment). Interview assessment seems therefore better able than self-report to identify inflated illusions about self. We may speculate that the interpersonal dynamics that characterize faceto-face interviews present a greater threat to fragile selfesteem and therefore result in greater self-enhancement and denigration of others. In contrast, the maintenance of “normality” is more easily preserved in the impersonal context of completing a self-report measure. Thus, the differential interpersonal dynamics of the two assessment conditions may influence the nature of the response. Robins and Beer (2001) suggested that self-enhancement may be a way to regulate self-esteem under threat, and that narcissism is a maladaptive way to achieve positive self-esteem, especially in the long term. Although we are not, of course, arguing for a complete overlap between dismissing attachment, self-enhancement, and narcissism, it may still be useful to explore their commonalities and to connect the study of attachment to other subfields in psychology that are also concerned with the interface among the self, emotional processing, and different approaches to the interpersonal world. For example, fragile self-esteem constitutes an important part of narcissism (Tracy & Robins, in press). Furthermore, self-enhancement can be explained by individual differences in narcissism (John & Robins, 1994). Robins

Gjerde et al. / ATTACHMENT, PERSONALITY, AND METHOD and Beer (2001) reported that self-enhancement was moderately related to narcissism (r = .36, p < .05). Preliminary analyses of our database also suggest that dismissing attachment is positively related to narcissism. Attempt to seek confirmation of their superiority may explain the tendency of dismissing individuals to seek close relationships with preoccupied individuals (Collins & Read, 1990). These relationships also may provide a context where the dismissing partner can experience control—a central aspect of how they regulate emotions in close relationships. These processes relate to the notion of “thick-skinned” narcissism, whereby “the individual deposits his or her own perceived inadequacies in others, whom he or she can then dismiss, denigrate, and devalue” (Fonegy, 1999, p. 607). We are currently conducting further examinations of the relations among dismissing attachment, narcissism, and self-enhancement. Scale Construction Strategies for Self-Reported Attachment To discriminate more reliably between self-report measured attachment patterns, the use of subtle items (i.e., items whose meaning is not immediately transparent to the respondent, not characterized by high social desirability, yet correlate well with the criterion; e.g., Block, 1965) deserves further examination. Furthermore, self-report attachment inventories might include response validity scales to recognize persons not well identified by a given scale (e.g., Tellegen, 1988). The strategy of correlating narrative qualities (e.g., coherence of discourse) with a large pool of self-report items also should be considered. We are currently using this method to develop a subtle self-report scale for dismissing attachment. In a particularly important study, Waters et al. (2002) reported that the new Anxiety and Avoidance scales (Brennan et al., 1998) were uncorrelated with the coherence scales of both the Adult Attachment Interview (AAI) and—especially noteworthy—the Close Relationship Interview (CRI). At the same time, the Anxiety and Avoidance scales were highly correlated with other self-report scales. Given that the Anxiety and Avoidance scales have been promoted as the new criterion scales for assessment of romantic attachment, the results presented by Waters and his colleagues—if replicated—cast serious doubt on the validity of the two scales advanced by Brennan et al. (1998). Avoidance belongs to the interpersonal domain and is not identical to intrapsychic defenses. We need to be mindful of this distinction if we want to clearly identify dismissing individuals. Caveats The stronger relations obtained between RAQ prototype scores relative to ECR prototype scores with person-

1413

ality attributes may partly reflect common method variance insofar as both the RAQ and the CAQ measures were based on observer data. It seems nonetheless unlikely that the relations between RAQ and CAQ personality attributes can be fully reduced to common method variance because (a) the RAI coders had no previous knowledge of the participants, (b) the RAQ items were very different from the CAQ items, (c) the semipartial correlations showed a differentiated and theory-consistent pattern, and (d) RAQ prototype scores were significantly associated with corresponding ECR self-report prototype scores and many other different self-report scales. In sum, the RAQ prototype scores did not seek their validation exclusively by referencing observer data. The interview coders worked directly from videotaped interviews. This procedure is not new, although Bartholomew (Bartholomew & Horowitz, 1991) coded her interviews from audiotapes rather than from videotapes. Videotaped interviews provide more information than audiotapes, but they also may leave coders more susceptible to the social stimuli qualities of the participants (e.g., physical attractiveness, mannerisms, and tone of voice) and thus distract the coders from important aspects of the participants’ discourse. Nonetheless, the coding procedure used in this study is an encouraging alternative to self-report measures without requiring the investment of time and money required by transcribing and then microcoding the interviews. However, coding attachment interviews from videotapes (or audiotapes) appears viable only when the goal is to identify the main attachment categories. Detailed evaluation of discourse qualities and reliable identification of certain types of coherence violations (e.g., passivity of speech and disorganized lapses in discourse) most likely require the use of transcripts (Main & Goldwyn, 1998). Examination of response biases in interviews needs further scrutiny. One might argue that interviewees filter statements to present themselves in a desirable light, engage in impression management, or are victims of selfdeception. Individuals may differ in the degree to which they have been taught that they can criticize others when talking to strangers. Securely attached individuals seem more willing to say negative things about their parents than are insecure “idealizing” individuals, but this proclivity may not be uniquely related to attachment (Thorne, personal communication, June 8, 2003). In contrast to self-report, however, self-enhancement may not be seen as a response bias per se in interview assessment but rather as an intrinsic part of an attachment pattern. In a similar vein, repressive defensiveness and interpersonal variables may also not be viewed as pure indicators of external validity as much as definitional components of attachment working models. Hence, it

1414

PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

might be more correct to view these analyses as contributing to the construct validity of the attachment prototype scores rather than to their external validity. Conclusion Although interview assessment was better than selfreport assessment in predicting observations of personality, whether this superiority is strong enough to justify the time-consuming and resource-demanding analysis of interviews, even when they are not transcribed, is a difficult issue. Based on the findings of this study, our previous results for the ECR prototype scores (Onishi et al., 2001), and the substantial body of theoretically meaningful results generated by self-report measures of romantic attachment, the use of the self-report method in the study of romantic attachment seems justified. Narrative (Waters, Rodrigues, & Ridgeway, 1998) and projective (George, West, & Pettem, 1997) procedures are now being introduced to measure attachment. Studying the convergence of results across different methods constitutes a major challenge for attachment research. We hope that the current research, its limitations notwithstanding, has contributed to this aim. NOTES 1. It should be noted that self-report scales contain items that dismissing individuals are likely to endorse because the content is both consciously accessible and acceptable, such as “I prefer not to be too close to romantic partners” (Brennan, Clark, & Shaver, 1998). Many dismissing individuals may be able to provide valid answers to such questions. It seems important to differentiate between domains where dismissing individuals can and cannot provide valid answers, that is, to determine the limits of their self-insight. 2. Note that the Social Desirability Scale (SDS) alpha reliability in the Crowell et al. (1996) study was, especially for a 33-item test, quite low (.58). This translates into an average interitem correlation of less than .05. Such a low alpha leads to substantial attenuation of relations with other variables. Bakermans-Kranenburg and van Ijzendorn (1993) used an abbreviated 10-item Dutch version of the SDS and reported no influence of social desirability on AAI responses. As Cliff (1987) noted, results are not dependable if the reliability of the control variable is low. Hence, the issue of response bias in interview assessment of attachment cannot be said to have been settled. 3. To convey the nomological network surrounding this measure, we present some of its self-report correlates. The RAQ Secure prototype scores were positively associated with introspection, achievement, positive affect, social closeness, and psychological well-being; the RAQ Preoccupied prototype scores were positively associated with depressive symptoms, shyness, negative affect, neuroticism, interpersonal dependency, and anxiety and negatively associated with self-esteem and ability to handle stress; the RAQ dismissing prototype scores were positively associated with defensive separation and negatively associated with introspection, openness to experience, and interpersonal dependency. Overall, these correlates seem consistent with attachment theory. 4. Of the 50 method-congruent secure individuals, 45 were involved in a serious relationship. Of the 12 individuals who described themselves as secure but were judged to be dismissing according to the interview, only 3 were seriously involved. This difference is statistically significant (χ2 = 3.87, p < .05). 5. To achieve a complete portrayal of the method-incongruent participants, it would be useful to compare these participants with individ-

uals who were seen as dismissing on both measures. Unfortunately, the sample was too small to accomplish this aim (only 5 individuals fell into this category). Low Ns also prevented comparisons of the other kind of method-incongruent individuals, for example, individuals who were evaluated as dismissing on self-report but were seen as secure according to the interview assessment. We also expect that the latter kind of method-incongruence would be less common than the former. Larger samples are needed to examine this issue. REFERENCES Bakermans-Kranenburg, M., & van Ijzendorn, M. (1993). A psychometric study of the Adult Attachment Interview: Reliability and discriminant validity. Developmental Psychology, 29, 870-879. Bartholomew, K., & Horowitz, L. (1991). Attachment styles among young adults: A test of a four-category model. Journal of Personality and Social Psychology, 61, 226-244. Bernier, A., & Dozier, M. (2002). Assessing adult attachment: Empirical sophistication and conceptual bases. Attachment and Human Development, 4, 171-179. Block, J. (1965). The challenge of response sets. New York: AppletonCentury-Crofts. Block, J. (1978). The Q-sort method in personality assessment and psychiatric research. Palo Alto, CA: Consulting Psychologists Press. (Original work published 1961) Block, J. H., & Block, J. (1980). The role of ego-control and egoresiliency in the organization of behavior. In W. A. Collins (Ed.), Minnesota symposia on child psychology (Vol. 13, pp. 39-101). Hillsdale, NJ: Lawrence Erlbaum. Bowlby, J. (1969). Attachment and loss: Vol. 1. Attachment. New York: Basic Books. Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-report measurement of adult attachment: An integrative overview. In J. A. Simpson & W. S. Rholes (Eds.), Attachment theory and close relationships (pp. 46-76). New York: Guilford. Brennan, K. A., & Shaver, P. R. (1990). Dimensions of adult attachment and the dynamics of romantic relationships. Unpublished manuscript, State University of New York, Buffalo. Carlson, K. S., & Gjerde, P. F. (1996). [The Romantic Attachment Qset (RAQ)]. Unpublished measure, University of California, Santa Cruz. Cliff, N. (1987). Analyzing multivariate data. San Diego, CA: Harcourt Brace Jovanovich. Cockery, L., & Gjerde, P. F. (1989). Romantic Attachment Interview. Unpublished manuscript, University of California, Berkeley. Collins, N. L., & Read, S. J. (1990). Adult attachment, working models, and relationship quality in dating couples. Journal of Personality and Social Psychology, 58, 644-663. Colvin, C. R., Block, J., & Funder, D. C. (1995). Overly positive selfevaluations and personality: Negative implications for mental health. Journal of Personality and Social Psychology, 68, 1152-1162. Cooper, M. L., Shaver, P. R., & Collins, N. L. (1998). Attachment styles, emotion regulation, and adjustment in adolescence. Journal of Personality and Social Psychology, 24, 1380-1397. Cronbach, L. J. (1955). Processes affecting scores on “understanding of others” and assumed similarity. Psychological Bulletin, 52, 177193. Crowell, J. A., Fraley, R. C., & Shaver, P. R. (1999). Measurement of individual differences in adolescent and adult attachment. In J. Cassidy & P. R. Shaver (Eds.), Handbook of attachment: Theory, research, and clinical applications (pp. 434-465). New York: Guilford. Crowell, J. A., Treboux, D., & Waters, E. (1999). The Adult Attachment Interview and the Relationship Questionnaire: Relations to reports of mothers and fathers. Personal Relationships, 6, 1-18. Crowell, J. A., Waters, E., Treboux, D., O’Connor, E., Colon-Downs, C., Feider, O., et al. (1996). Discriminant validity of the Adult Attachment Interview. Child Development, 67, 2584-2599. Crowne, D. P., & Marlowe, D. (1964). A new scale of social desirability independent of psychopathology. Journal of Consulting Psychology, 24, 349-354.

Gjerde et al. / ATTACHMENT, PERSONALITY, AND METHOD DeHaas, M. A., Bakermans-Kranenburg, M., & van Ijzendorn, M. H. (1994). The Adult Attachment Interview and questionnaires for attachment style, temperament, and memories of parental behavior. Journal of Genetic Psychology, 155, 471-486. Dozier, M., & Kobak, R. R. (1992). Psychophysiology in attachment interviews: Converging evidence for deactivating strategies. Child Development, 63, 1473-1480. Feeney, J. A., & Collins, N. L. (2001). Predictors on caregiving in adult intimate relationships: An attachment theoretical perspective. Journal of Personality and Social Psychology, 80, 972-994. Feeney, J. A., Noller, P., & Hanrahan, M. (1994). Assessing adult attachment. In M. B. Sperling & W. H. Berman (Eds.), Attachment in adults: Clinical and developmental perspectives (pp. 128-152). New York: Guilford. Fonegy, P. (1999). Psychoanalytic theory from the viewpoint of attachment. In J. Cassidy & P. R. Shaver (Eds.), Handbook of attachment: Theory, research, and clinical applications (pp. 595-624). New York: Guilford. Fraley, R. C., & Shaver, P. R. (1997). Adult attachment and the suppression of unwanted thoughts. Journal of Personality and Social Psychology, 73, 1080-1091. Furman, W., Simon, V. A., Shaffer, L., & Bouchey, H. A. (2002). Adolescents’ working models and styles of relationships with parents, friends, and romantic partners. Child Development, 73, 241-255. George, C., Kaplan, N., & Main, M. (1995). Adult Attachment Interview. Unpublished manuscript, University of California, Berkeley. (Original work published 1985) George, C., & Solomon, J. (1999). Attachment and caregiving: The caregiving behavioral system. In J. Cassidy & P. R. Shaver (Eds.), Handbook of attachment: Theory, research, and clinical applications (pp. 649-670). New York: Guilford. George, C., & West, M. (1999). Developmental vs. social personality models of adult attachment and mental health. British Journal of Medical Psychology, 72, 285-303. George, C., West, M., & Pettem, O. (1997). [The Adult Attachment Projective Test]. Unpublished attachment measure and coding manual, Mills College. Gjerde, P. F. (1995). Alternative pathways to chronic depressive symptoms in young adults: Gender differences in developmental trajectories. Child Development, 66, 1277-1300. Griffin, D., & Bartholomew, K. (1994). Models of the self and other: Fundamental dimensions underlying measures of adult attachment. Journal of Personality and Social Psychology, 67, 430-445. Hazan, C., & Shaver, P. R. (1987). Romantic love conceptualized as an attachment process. Journal of Personality and Social Psychology, 52, 511-524. Hazan, C., & Shaver, P. R. (1990). Experiences in Close Relationship Scale. Unpublished manuscript, Cornell University. Hesse, E. (1999). The adult attachment interview: Historical and current perspectives. In J. Cassidy & P. R. Shaver (Eds.), Handbook of attachment: Theory, research, and clinical applications (pp. 395-433). New York: Guilford. John, O. P., & Robins, R. (1994). Accuracy and bias in self-perceptions: Individual differences in self-enhancement and the role of narcissism. Personality and Social Psychology, 66, 206-219. Klohnen, E., & Bera, S. (1998). Behavioral and experiential patterns of avoidantly and securely attached women across adulthood: A 32-year longitudinal perspective. Journal of Personality and Social Psychology, 74, 211-223. Kobak, R. R., & Sceery, A. (1988). Attachment in late adolescence: Working models of attachment, affect regulation, and perceptions of self and others. Child Development, 88, 135-146.

1415

Main, M. (1991). Metacognitive knowledge, metacognitive monitoring, and singular (coherent) vs. multiple (incoherence model of attachment: Findings and direction for future research. In C. M. Parkes, J. Stevenson-Hinde, & P. Marris (Eds.), Attachment across the life cycle (pp. 127-159). London: Tavistock/Routledge. Main, M., & Goldwyn, R. (1998). Adult attachment classification system. Unpublished manuscript, University of California, Berkeley, Department of Psychology. McNemar, Q. (1969). Psychological statistics (4th ed.). New York: John Wiley. Mickelson, K. D., Kessler, R. C., & Shaver, P. R. (1997). Adult attachment in a nationally representative sample. Journal of Personality & Social Psychology, 73, 1092-1106. Mikulincer, M., & Orbach, I. (1995). Attachment styles and repressive defensiveness: The accessibility and architecture of affective memories. Journal of Personality and Social Psychology, 68, 917-925. Onishi, M., Gjerde, P. F., & Block, J. (2001). Personality implications of romantic attachment patterns in young adults: A multimethod, multi-informant study. Personality and Social Psychology Bulletin, 27, 1097-1110. Ozer, D. J., & Reise, S. P. (1994). Personality assessment. Annual Review of Psychology, 45, 357-388. Paulhus, D. L. (1987). Two-component model of socially desirable responding. Journal of Personality and Social Psychology, 46, 598-609. Robins, R. W., & Beer, J. S. (2001). Positive illusions about the self: Short-term benefits and long-term costs. Journal of Personality and Social Psychology, 80, 340-352. Rorer, L. G. (1990). Personality assessment: A conceptual survey. In L. A. Pervin (Ed.), Handbook of personality (1st ed., pp. 693-722). New York: Guilford. Shaver, P. R., Belsky, J., & Brennan, K. A. (2000). Comparing measures of adult attachment: An examination of interview and selfreport methods. Personal Relationships, 7, 25-43. Shaver, P. R., & Brennan, K. A. (1992). Attachment styles and the “Big Five” personality traits: Their connections with each other and with romantic relationship outcomes. Personality and Social Psychology Bulletin, 18, 536-545. Shaver, P. R., & Mikulincer, M. (in press). What do self-report attachment measures assess? In W. S. Rholes & J. A. Simpson (Eds.), Adult attachment: Theory, research, and clinical implications. New York: Guilford. Siegrist, M. (1996). The influence of self-consciousness on the internal consistency of different scales. Personality and Individual Differences, 20, 115-117. Tellegen, A. (1988). The analysis of consistency in personality assessment. Journal of Personality, 56, 621-663. Waters, E., Crowell, J., Elliot, M., Corcoran, D., & Treboux, D. (2002). Bowlby’s secure base theory and the social/personality psychology of attachment styles: Work(s) in progress. Attachment and Human Development, 4, 230-242. Waters, H. S., Rodrigues, L. M., & Ridgeway. D. (1998). Cognitive underpinnings of narrative attachment assessment. Journal of Experimental Child Psychology, 71, 211-234. Wiggins, J. (1973). Personality and prediction: Principles of personality assessment. Reading, MA: Addison-Wesley. Zimmermann, P., Maier, M. A., Winter, M., & Grossmann, K. E. (2001). Attachment and adolescents’ emotion regulation during a joint problem solving task with a friend. International Journal of Behavioral Development, 25, 331-343. Received January 24, 2003 Revision accepted January 26, 2004