Académique Documents
Professionnel Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: For over two decades, individual differences in adult attachment style have been conceptualized and
Available online 27 December 2008 measured in terms of anxiety, avoidance, and security. During this time, the prevailing assumption has
been that an adults attachment style is a relatively stable disposition, rooted in internal working models
Keywords: of self and relationship partners. These models are based on previous experiences in close relationships.
Attachment Recent research, however, suggests that levels of attachment anxiety, avoidance, and security are also
State affected by situational factors. To capture temporary uctuations in the sense of attachment security
Trait
and insecurity we developed a state adult attachment measure (SAAM). Exploratory and conrmatory
Measurement
factor analyses yielded three reliable subscales measuring state levels of attachment-related anxiety,
avoidance, and security. Additional studies demonstrated both convergent and discriminant validity of
the new measure, and its sensitivity to a variety of experimental manipulations. Our discussion focuses
on potential uses for the SAAM for both researchers and clinicians.
2008 Elsevier Inc. All rights reserved.
0092-6566/$ - see front matter 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jrp.2008.12.009
O. Gillath et al. / Journal of Research in Personality 43 (2009) 362373 363
behavioral systems. According to Bowlby (1969/1982), mental rep- 3. The need for a state measure of attachment style
resentations of self and other are largely accurate reections of ac-
tual experiences, especially in the context of close relationships, Based on the conceptualizations and research reviewed above,
which are revised and updated as a person enters new relation- it seems clear that atleast life events (especially those pertaining
ships and has new experiences (although Bowlby posited that to close relationships), and experimental activation of close rela-
working models that are formed early in life, or are particularly tionship schemas, temporarily affect peoples attachment style or
strong, will tend to persist). levels of security and insecurity. Moreover, these temporarily uc-
Numerous studies have provided evidence to support Bowl- tuations are not simply noise but result in meaningful behavior;
bys conceptualization of working models and the dynamic priming studies have shown that enhancing security or insecurity
attachment system. For example, Feeney and Noller (1992) inuences prosocial behavior such as volunteering, well-being
found when following a sample of young adults for 10 weeks and mental health (e.g., lowering PTSD symptoms), and increases
that the formation of a steady relationship was related to an in- intergroup tolerance (e.g., Mikulincer, Shaver, Gillath, & Nitzberg,
crease in attachment security and a decrease in attachment inse- 2005; for reviews see Gillath et al., 2008; Mikulincer & Shaver,
curity. Similarly, Hammond and Fletcher (1991) found that 2007a). In other words, attachment style uctuates across mo-
involvement in satisfying relationships at one point in time ments in time and situations, albeit within a range constrained in
was associated with increased security later. Kirkpatrick and Ha- part by a persons stable dispositions. Fleeson (2001) has already
zan (1994) also found that relationship experiences moderate demonstrated similar ndings for the Big Five personality traits
the stability of attachment style when they followed college stu- although individuals average level of behavior relevant to each
dents over 4 years. Relationship breakups were associated with trait was very stable from one week to the other, their scores on
changes from secure to insecure attachment styles, and avoidant Big Five dimensions varied across specic occasions. Thus, scores
subjects who formed new relationships were less likely to re- on a state attachment measure have the potential to account for
main avoidant than those who did not. Likewise, Davila and Sar- a unique portion of the variance in psychological, behavioral, and
gent (2003) found that perceptions of greater interpersonal loss relationship outcomes (e.g., emotions, interpersonal interactions,
related to life events were positively associated with greater and relationship stability and satisfaction).
attachment insecurity on a day-to-day basis. Especially impor- This suggests that a measure enabling the detection of short-
tant for the current work, they also found that trait levels of term uctuations in attachment would benet researchers inter-
attachment security did not moderate this association, suggest- ested in attachment-related psychological and relationship pro-
ing that state attachment or uctuations in the sense of secu- cesses. Additionally, because more therapists are integrating
rity is independent from dispositional attachment style. attachment theory into their practices (e.g., Hughes, 2007; Johnson
Despite their encouraging preliminary ndings, Davila and Sar- & Whiffen, 2003; Oppenheim & Goldsmith, 2007), measuring the
gent also found high week-to-week correlations between the effect of therapy on attachment style, and specically the sense
attachment scores (over .90), which means they may not have of security, is essential; therefore, a state measure would be bene-
adequately captured the state component of attachment. This cial for clinicians as well.
highlights the need for a scale that is solely dedicated to the Despite the apparent importance of momentary uctuations in
measurement of such a construct. attachment style, there is currently no way to measure such state
Recent empirical ndings provide further evidence that changes. The most widely used measure of adult attachment style,
working models, and thus attachment styles, are malleable the experiences in close relationships scale (ECR; Brennan, Clark, &
across even very short durations, such as minutes or hours. Shaver, 1998) is a highly reliable and valid instrument, explicitly
For instance, reminding people of times when they have felt se- designed to measure stable dispositions. Accordingly, it instructs
cure, anxious, or avoidant momentarily activates a specic participants to reect on their general experiences in relationships
attachment schema (Baldwin, Fehr, Keedian, Seidel, & Thomson, (either with a specic gure or all close gures), which probably
1993; Baldwin, Keelan, Fehr, Enns, & Koh-Rangarajoo, 1996). activates generalized, abstracted working models at the expense
This momentary activation of a schema seems to temporarily of momentary models. Moreover, because the ECR items were
override stable dispositions in terms of inuencing participants written to reect generalized close relationship feelings, the scale
perceptions, expectations, and behaviors. For example, Gillath cannot be easily converted to a state measure (e.g., by changing
and Shaver (2007) showed that priming a specic attachment the instructions or altering the wording of individual items). We
context (by asking people to imagine their partner as either have therefore developed the state adult attachment measure
responsive and sensitive or unresponsive and insensitive) affects (SAAM). In the current article, we describe the development of
their responses to relationship-related scenarios, and Gillath the SAAM and several studies that tested its convergent, discrimi-
et al. (2006) showed that such priming also affects participants nant, and construct validity.
goal pursuits. Gillath et al. (2008) bring evidence to suggest
that these temporary changes can persist for a relatively long 3.1. What should a state attachment measure look like?
time.
According to Baldwin and others, changes in attachment style Currently, there are two main approaches to measuring adult
are possible because people simultaneously hold in their minds attachment style: self-reports and interviews.1 The Adult Attach-
several models of self and other ordered in a hierarchy (Baldwin ment Interview (AAI; George, Kaplan, & Main, 1985) and measures
et al., 1996; Collins & Read, 1994; Klohnen, Weller, Luo, & Choe, based on it (e.g., Bartholomew & Horowitz, 1991) usually denes
2005; Pierce & Lydon, 2001). At higher levels of organization, these attachment in terms of three or four categories; however several
models include abstract rules or assumptions about attachment of the self-report measures, including the ECR (and the revised
relationships, and at lower levels, they include information about ECR-R; Fraley, Waller, & Brennan, 2000) conceptualize adult attach-
specic relationships and events within relationships (e.g., Collins ment in terms of two orthogonal dimensions (anxiety and avoid-
& Read, 1990; Overall, Fletcher, & Friesen, 2003). Each of these ance), with security dened as low scores on both of these
models can be activated and made more accessible at any given
moment. Levels of attachment anxiety, avoidance, and security 1
There are other ways to measure attachment style such as projective measures;
are then set as a function of the model or schema that is most however, self-reports and interviews are the most common ones according to
strongly activated. Mikulincer and Shavers (2007a) literature review.
364 O. Gillath et al. / Journal of Research in Personality 43 (2009) 362373
dimensions. Because of this, one might assume that a state measure Second, in order to cover the entire range of thoughts, feelings,
would have a similar structure of two orthogonal dimensions. Nev- and behaviors related to attachment, we wrote several items to
ertheless, other measures, such as Simpsons (1990), and Hazan represent the dimensions formed by the interaction of anxiety
and Shavers (1987) original measure have suggested a different and avoidance (i.e., secure vs. fearful, and preoccupied vs. dismiss-
structure in which security is an independent dimension, related ing). We also wrote additional items to reect aspects of attach-
negatively, but moderately, to anxiety and avoidance (which are in ment styles that are underrepresented on current measures, such
turn mostly or totally independent of each other). as the low end of anxiety (e.g., I feel relaxed knowing that close
Levy and Davis (1988) also found a three-factor structure, others are there for me right now), which is represented by only
where scores on security were moderately negatively correlated a single item in the ECR.
with scores on avoidant attachment, and weakly negatively corre- Third, we invited several attachment researchers and a team of
lated with scores on attachment anxiety (there was a near-zero graduate students familiar with attachment to examine the rewrit-
correlation between anxiety and avoidance). Carver (1997), ten and new items and comment on their readability, face validity,
attempting to overcome previous measures limitations in assess- and comprehensiveness, and to make suggestions for additional
ing the sense of security, developed a similarly structured adult items that could be included. These three processes resulted in
attachment measure. Specically aiming to measure security sep- 52 preliminary items.
arately from low insecurity, Carver conceptualized security as Although many of the items were directly written to emphasize
reecting an appreciation of ones relationship as a safe haven a state-focused context, we wanted to further ensure that partici-
and a secure base for exploration, which resulted in three scales: pants would respond to the items in terms of how they were cur-
avoidance, ambivalence (reecting two facets of anxiety: worry rently feeling. Thus, we included the following instructions (similar
and desire for merger), and security. Carvers Security scale was re- to other state measures):
lated inversely to the Avoidance scale, and was relatively unrelated The following statements concern how you feel right now.
to either of the Ambivalence scales. Carvers Avoidance scale was Please respond to each statement by indicating how much you
also relatively unrelated to either Ambivalence scales. agree or disagree with it as it reects your current feelings. Please
This evidence suggests that an alternative model for the struc- circle the number on the 17 scale that best indicates how you feel
ture of adult attachment style may include a separate security at the moment.
scale, or perhaps two dimensions running diagonally (as compared
to the common anxiety avoidance matrix), one from security to 4.1. Method
insecurity, and the other from high anxiety and low avoidance to
high avoidance and low anxiety (e.g., Asendorpf, Banse, Wilpers, 4.1.1. Participants
& Neyer, 1997; Banse, 2004; Elizur & Mintzer, 2003; Mikulincer We administered the preliminary measure to three hundred
& Shaver, 2007b). This idea of rotated dimensions ts with Kobak forty seven undergraduate students (258 women and 89 men,
et al.s work (Kobak, Cole, Ferenz-Gillies, Fleming, & Gamble, ranging from 18 to 35 years of age, Mdn = 19) as part of a longer
1993), Fyffe and Waterss (1997) two-dimensional scoring system survey (see Study 4). Two hundred and twenty eight had never
for the AAI, and Mikulincer and Shavers model (2003, 2007b) of been married, ve married, one divorced, and one hundred thirteen
attachment processes. Based on Mikulincer and Shavers (2003) indicated other or declined to say. One hundred and ninety par-
dynamic model, which includes three modes of the attachment ticipants reported being currently in a romantic relationship (54%).
system above and beyond a persons location on the two-dimen- Participants responded to each SAAM item using a 7-point scale,
sional space, a three-factor model may adequately capture changes ranging from 1 (Disagree strongly) to 7 (Agree strongly), with 4 (neu-
in each one of these possible dimensions avoidance, anxiety, and tral/mixed) as the midpoint of the scale. The 52 items were pre-
security. sented in a randomized order.
Still, one cannot reject alternative structures for a state attach-
ment measure. We therefore tested various possible structures, 4.1.2. Exploratory factor analysis
with a varying number of dimensions, and different links between We conducted an exploratory factor analysis testing 1, 2, 3, and
the dimensions. The only restriction we had for our measure was 4-factor solutions, which reected common structures in existing
that the items had to reect attachment-related thoughts, feelings, trait attachment style measures. A one-factor solution might indi-
or behaviors that would be interpretable in light of existing mea- cate a single secureinsecure bipolar dimension, which we thought
sures and structural conceptualizations. In the next section, we de- possible but unlikely for state-based attachment. A two-factor
scribe our rst study in which we generated items for the SAAM solution might suggest a similar structure to Brennan et al.s
and conducted an exploratory factor analysis to determine the (1998) ECR, with a state dimension for both anxiety and avoidance.
optimal structure for a state measure of attachment. A three-factor structure could represent an organization similar to
Hazan and Shavers (1987) prototype approach of secure, avoidant,
and anxiousambivalent styles or Collins and Reads (1990) dimen-
4. Study 1 sions of anxiety, closeness, and dependency. A four-factor structure
could represent the quadrants formed by the RQ (Bartholomew &
To develop a measure that could tap into momentary feelings of Horowitz, 1991).
attachment, we rst revisited the 36-item ECR (Brennan et al., We predicted different patterns of correlations between the fac-
1998), which itself resulted from factor analyses of many of the tors as a function of the different factor solutions. For example, we
earlier self-report measures of attachment. We reworded each predicted that with a two-factor solution that was similar to Bren-
ECR item that could be transformed to emphasize a state context. nan et al.s orthogonal structure of attachment, we might observe
For example, we rewrote I get uncomfortable when a romantic relatively uncorrelated factors. However, a two-factor structure
partner wants to be very close as I would be uncomfortable hav- that contrasted secure vs. insecure styles, or approach-related vs.
ing a good friend or a relationship partner close to me now. We avoidance-related styles, might be negatively correlated. A three-
added the term good friend to allow people with no romantic part- factor structure similar to Hazan and Shaver (1987) might result
ner to also complete the questionnaire without skipping items. (In in a relatively small negative correlation between the secure and
the nal version the word now was dropped from this item and anxious factors, but larger, negative correlations between the other
remained only in the general instructions of the questionnaire). two-factor pairings (e.g., Shaver & Brennan, 1992). A four-factor
O. Gillath et al. / Journal of Research in Personality 43 (2009) 362373 365
model that maps onto the four categories provided by the 2- than .40 on the proposed factor5, resulting in a 21-item scale, with 7
dimensional space created by the ECR dimensions would have pre- items measuring each of the three content domains. Table 1 shows
dictable associations according to a circumplex-like model: adjoin- the loadings of the 21 items on each of the factors. The security
ing categories would be relatively uncorrelated whereas opposing and anxiety factors were correlated .09, the security and avoidance
categories would be strongly negatively correlated. factors were correlated .50 and the anxiety and avoidance factor
were correlated .02. Table 2 includes psychometric characteristics
4.2. Results and discussion of the measure: alpha reliabilities and inter-correlations between
the scales. Cronbach alpha coefcients for the three SAAM subscales
We examined the underlying structure of the 52 items through ranged from .83 to .87, suggesting good internal consistency.
a series of exploratory factor analyses in SAS 9.1.3 (20022004)
with unweighted least squares extraction in the initial factoring 5. Study 2
process, and Promax rotation, because we did not expect that the
factors would necessarily be orthogonal.2 We examined the initial The purpose of study 2 was to conrm the factor structure of
52-item solution using the scree test (Cattell, 1966; 1978, p. 7), the SAAM using conrmatory factor analysis (CFA) and a different
rather than the traditional Eigenvalue greater than one test.3 Using sample. In a CFA, a factor structure is explicitly hypothesized and
the scree test, the rst signicant jump between eigenvalues seemed is tested for its t with the observed covariance structure of the
to be between the fourth (eigenvalue = 1.51) and the third factor measured variables. The approach also allows for testing the rela-
(eigenvalue = 2.38), or more conservatively, between the third factor tive t of competing factor models, with aid of theory, modication
and the second factor (eigenvalue = 7.37). Therefore, the scree pat- indexes, and the pattern and signicance of factor loadings. Based
tern suggested either a three- or two-factor solution, but because on our exploratory factor analysis results, we tested the t of one-
we had theoretical ideas about one-factor to four-factor solutions, factor, two-factor, three-factor, and four-factor solutions with a
we decided to examine these solutions as well. new set of data.
In the four-factor solution only two items loaded substantially
on the fourth factor (0.69 and 0.68), creating a doubleton or non- 5.1. Method
factor (e.g., Gorsuch, 1983). The two items, I dont need anyone
to help me and I dont need anybodys help or support, sug- 5.1.1. Participants
gested that the way participants were responding to the help-re- We administered the 21-item SAAM to 1433 undergraduate
lated content of the items was forcing them into a fourth factor students (939 women and 494 men, ranging from 18 to 48 years
which was less related to a style of attachment than to need for of age, Mdn = 19) as part of a longer online survey. Of the sample,
help. Therefore, we examined the three-factor solution. This solu- 46% were single, 16% were casually dating, 37% were in a relation-
tion included three factors with items that had high positive load- ship, and 1% were married. Of the participants, 37% were involved
ings on their respective factors, no negative main loadings on any in a romantic relationship. As with the EFA sample, participants re-
of the factors,4 and negligible cross-loadings on other factors. For sponded to each SAAM item using a 7-point scale, ranging from 1
example, there were many loadings above .50: the rst factor had (Disagree strongly) to 7 (Agree strongly), with 4 (neutral/mixed) as
14 loadings, the second factor had 12 loadings, and the third factor the midpoint of the scale. The 21 items were presented in a ran-
had 8 loadings above this value (with three other items loading just domized order.
under .50). In addition, there were several items that did not load
strongly on any of the factors. The two-factor solution also had a 5.1.2. Conrmatory factor analyses
number of high loadings but they were relatively unbalanced (the All models examined the structural relation of the 21 items se-
rst factor had 23 loadings above .50 whereas the second factor lected as the result of the EFA. The one-factor model consisted of a
had just 14). factor ranging from positive to negative feelings and thoughts
In the next step, we reduced the number of items from 52 to 32, about attachment. We tested two two-factor models, one with a
to include only the items that had high loadings on both the two- security factor and an insecurity factor (subsuming avoidance
and three-factor solutions, and then once again we examined the and anxiety), and one with an approach factor (subsuming security
factor structure using exploratory factor analyses. This analysis and anxiety) and an avoidance factor. The three-factor model in-
suggested that the best structure was a 3-factor solution, which in- cluded subscales of anxiety, avoidance, and security. Lastly, the
cluded factors relating to security, avoidance, and anxiety, four-factor model was similar to the three-factor model except
accounting for 47.3% of the total variance (24.2%, 15.5%, and 7.6% that it differentiated between fearful avoidance and dismissing
for the three factors, respectively). avoidance. Using M Plus 4.0 (Muthen & Muthen, 2006), maximum
The nal version of the scale was obtained through several l- likelihood CFAs of the SAAM scores were conducted comparing the
tering steps, where we eliminated items that had cross-loadings one-, two-, three-, and four-factor competing rst-order models.
greater than .20 on more than one factor or had loadings smaller
5.2. Results and discussion
2
We also tested orthogonal rotations. The results were not substantially different.
3
We did not employ the eigenvalue-greater-than-one test (e.g., Kaiser, 1960), 5.2.1. One-factor model
because by using an arbitrary cut-off it necessitates excluding a factor with an The one-factor model did not provide a good t to the data:
eigenvalue close to 1.00 (factor six: eigenvalue = .95) while including a factor not v2(189, N = 1433) = 6397.45; CFI = .59; RMSEA = .15.
much higher than 1.00 (factor ve: eigenvalue = 1.20), even though the two factors
are relatively close together in size (for other problems with the eigenvalue > 1 test,
see Lee and Ashton, 2007). 5.2.2. Two-factor models
4
Although there were no items with sizable negative loadings on the anxiety and Although the two-factor solutions resulted in improved t, nei-
avoidance factors in the three factor solution, there were a few items that negatively ther of the two-factor models provided an adequate t to the data.
loaded on the security factor (although most of them had sizable cross-loadings on
other factors). Using CFA, which we describe later in the current section, we tested a
5
three-factor model that incorporated the best items that negatively loaded on We included a seventh item for avoidance that was just below the .40 primary
security, but we found that the unipolar security factor contributed to a better tting loading cut-off (i.e., .393) because it tapped into the avoidance qualities of self-
model (v2(186, N = 1433) = 1647.25; CFI = .90; RMSEA = .07) than the bipolar security reliance and tough-minded independence (cf. 6667; Brennan et al., 1998) that
factor (v2(186, N = 1433) = 2191.58; CFI = .87; RMSEA = .09). were not measured with the other six items.
366 O. Gillath et al. / Journal of Research in Personality 43 (2009) 362373
Table 1
The nal set of SAAM items and their loadings as part of the exploratory factor analysis.
Note: The table above lists items by their original order in the 52 preliminary item set. To employ the SAAM in research, randomize the order of items or email the authors for
a formatted version.
a
The word emotionally was added in later versions of the scale for clarity.
Table 2
SAAM scale inter-correlations and internal consistency reliabilities across studies.
N Inter-correlations Reliabilities
Security with anxiety Security with avoidance Anxiety with avoidance Security Anxiety Avoidance
Study 1/4 347 .09 .48* .02 .87 .84 .83
Study 2/3 1433 .02 .61* .09 .91 .85 .87
Study 5 60 .28** .56* .28** .90 .81 .85
Study 6 142 .10 .37* .18* .82 .83 .71
Study 7 345 .02 .42** .07 .86 .83 .77
*
p<.05
**
p<.01.
The security-insecurity two-factor model did not provide a good t Overall, Study 2 results suggest that a three-factor model appro-
to the data: v2(188, N = 1433) = 4875.67; CFI = .69; RMSEA = .13. priately describes the structure of state attachment. Next we
Similarly, the approachavoidance two-factor model did not pro- turned to examining the SAAMs stability.6
vide a good t to the data either: v2(188, N = 1433) = 4874.19;
CFI = .69; RMSEA = .13. In both models, the two factors were nega- 6. Study 3
tively correlated rather strongly (both rs = .68). However, the
amount of shared variance (46%) was not large enough to indicate A subsample of 93 participants from Study 2 (27 men and 66
the factors were opposites. women, ranging in age from 18 and 28, Mdn = 20), completed the
SAAM again three months later to examine testretest stability.
5.2.3. Three-factor model All SAAM subscales showed reasonable (and comparable) stability
The three-factor model provided a fairly good t to the data: over a period of three months (.51, .53, and .59 for SAAM anxiety,
v2(186, N = 1433) = 1647.25; CFI = .90; RMSEA = .07. Ideally, the avoidance, and security, respectively, all ps < .01). These stability
CFI should be higher than the .90 cut-off often specied (e.g., Jre- scores are a little higher but comparable to what one nds with
skog & Srbom, 1996), so we thought it would be appropriate to the state PANAS (Watson, Clark, & Tellegen, 1988), but less than
test a four-factor model to see if it would provide even better t. what is typically found with trait measures such as the sensitivity
to reward and punishment scale (Torrubia, vila, Molt, & Caseras,
5.2.4. Four-factor model
The securityanxietydismissingfearful four-factor model also
6
Two-factor analyses (using Promax rotation) were performed separately on the
provided a fairly good t to the data: v2(183, N = 1433) = 1526.01;
men and the women in the sample and revealed extremely similar structures for the
CFI = .91; RMSEA = .07. Although the model had a marginally better two groups. Similarly all the priming studies revealed no main effects or interactions
t than the three-factor model, the structure was redundant the for gender, except study 7 (women tended to report higher scores [M = 4.20] than
two avoidance factors were extremely highly correlated; r = .95 men [M = 3.75] on SAAM anxiety, and on SAAM security [M = 5.76 and M = 5.36,
so in the interest of parsimony, we selected the three-factor model. respectively]). Because most of our studies did not reveal main effects or interactions
with gender, and because gender does not usually play a role in attachment research,
The security and anxiety scales were correlated .02, the security
to simplify the presentation of results we decided not to include gender in the
and avoidance scales were correlated .61, and the anxiety and analyses of the current paper; nevertheless, future research should systematically
avoidance scales were correlated .09. examine reactions on the SAAM as a function of gender.
O. Gillath et al. / Journal of Research in Personality 43 (2009) 362373 367
2001; r = .87.89) or the Big Five (Watson, 2004; r = .79.89). Study ent feelings and emotions. Read each item and then indicate to
3 provided further evidence for the reliability of the SAAM, posi- what extent you feel this way right now, that is, at the present mo-
tioning it in an ideal place to capture temporary uctuations. Next ment, by writing a number from 1 (very slightly or not at all) to 5
we turned to examining the SAAMs validity. (extremely), in the space next to the item. Cronbachs alphas for
the positive affect and negative affect subscales in the current
7. Study 4 study were high (a = .92 and a = .81, respectively).
We also tested the extent to which the SAAM measures a con-
Studies 1 and 2 provided support for the triarchical structure of struct that is different from general anxiety using the state sub-
the SAAM and its internal reliability. However, before a self-report scale of the State-Trait Anxiety Inventory (STAI; Spielberger,
measure like the SAAM can be used with condence, it must dem- 1983). This 20-item instrument measures transient levels of anxi-
onstrate adequate validity as well. Therefore, Study 4 was designed ety and has strong psychometric properties (Spielberger, 1983).
to test the SAAMs convergent and discriminant validity. Conver- Participants are instructed to rate how well each item describes
gent validity refers to the extent to which the same construct is the way they feel right now using a 4-point scale, ranging from 1
measured by different methods or measures, whereas discriminant (not at all) to 4 (very much). Final scores are generated by summing
validity is dened as the extent to which the construct under all the items, and can range between 20 and 80. Cronbachs alpha
investigation is distinct from other constructs (Campbell & Fiske, in the current study was high (a = .92).7
1959). Additionally we used three trait measures to examine the
SAAMs convergent and discriminant validity. Rejection sensitivity
7.1. Method (RSQ; Downey & Feldman, 1996) was used to control for general
worries about rejection. This is an 18-item measure, asking people
7.1.1. Participants to indicate their degree of concern or anxiety about the outcome of
Study 4 used subsamples of the participants from Study 1. Par- various situations involving interpersonal rejection (e.g., How
ticipants were a random group of participants who took the concerned or anxious would you be over whether or not your
departmental screening questionnaire three months after they par- friend would want to help you out?) on a 6-point scale ranging
ticipated in the rst study. In each of the tables in which we report from 1 (very unconcerned) to 6 (very concerned). Alpha for the 18
results, we indicate the size of the subsample analyzed. The SAAM items was high (a = .90).
was completed as a part of a broader battery of questionnaires that To examine whether the SAAM captures feelings beyond rela-
included the measures described below. tionship satisfaction, we used the Perceived Relationship Quality
Component inventory (PRQC; Fletcher, Simpson, & Thomas,
7.1.2. Convergent validity 2000). The Inventory consists of 18 items, which can be divided
To assess convergent validity two attachment measures were into six components tapping perceived relationship quality. Partic-
used (1) the Experiences in Close Relationships questionnaire ipants are instructed to rate how well each item represents their
(ECR; Brennan et al., 1998), a 36-item self-report instrument that current partner and relationship using a 7-point Likert-type scale
measures attachment-related anxiety and avoidance. Participants ranging from 1 (not at all) to 7 (extremely). The overall alpha for
were asked to think about their general experiences in close rela- the 18 items was high (a = .94), therefore, a total score was calcu-
tionships (i.e., without focusing on a specic partner), and rate lated for each participant by averaging the 18 items. A higher score
the extent to which each item accurately described their feelings meant higher relationship satisfaction.
in such relationships using a 7-point scale ranging from 1 (not at Finally, because trust plays a major role in the way avoidantly
all) to 7 (very much). Eighteen items measured attachment-related attached people perceive their relationships, we decided to assess
anxiety (e.g., I worry about being abandoned) and 18 items mea- relationship trust using a modied version of Rempel and Holmes
sured avoidance (e.g., I prefer not to show a partner how I feel (1986) and Rempel, Ross, and Holmes (2001) trust scale. On this
deep down). The reliability and validity of the two subscales have 18-item scale participants rate their agreement with trust-related
been repeatedly demonstrated (e.g., Mikulincer & Shaver, 2007b); sentences on a 7-point scale ranging from 1 (strongly disagree) to
in the present study, alphas for anxiety and avoidance were .94 7 (strongly agree). In the current sample the alpha coefcient for
and .93, respectively. Anxiety and avoidance scores were moder- the total scale was high (a = .92), so we calculated a total score
ately correlated (r = .32, p < .05). (2) the Relationship Questionnaire for each participants by averaging all the items; a higher score
(RQ; Bartholomew & Horowitz, 1991), is a single item measure meant higher sense of trust in ones partner and relationship.
made up of four short paragraphs, each describing a prototypical
attachment pattern as it applies in close adult relationships (se- 7.2. Results and discussion
cure, fearful, preoccupied, and dismissing). Participants are asked
to rate how well each prototype describes them, using a 7-point The results section is divided into two subsections covering
scale ranging from 1 (not at all like me) to 7 (very much like me). convergent validity and discriminant validity.
Scores represent four continuous ratings based on each of the four
attachment patterns. Additionally, we asked participants to choose 7.2.1. Reliability
a single, best tting attachment pattern, a method which has been All the SAAM subscales had adequate reliability ranging from
shown useful in the past to enable researchers to assign partici- .83 to .87. As expected and as shown in Studies 1 and 2 the corre-
pants into a category.
7
Attachment security is frequently compared with high sense of self-esteem (e.g.,
7.1.3. Discriminant validity
Hart, Shaver, & Goldenberg, 2005), so to examine the possibility that the SAAM
We used three state measures to help us examine the SAAMs captures changes in state self-esteem rather than attachment security and insecurity,
discriminant validity. To rule out the possibility that the SAAM is we used four items from the STAI (Spielberger, 1983), which highly resemble items on
merely measuring mood uctuations we used the state form of the Heatherton and Polivys (1991) state self-esteem measure. The items were: I feel
the Positive and Negative Affect Schedule (PANAS; Watson et al., satised with myself, I feel like a failure (R), I lack self-condence (R), and I feel
inadequate (R). Alpha for the four items was adequate (a = .80). The new measure of
1988). This is a 20-item measure, assessing the way people feel state self-esteem score was as expected only moderately correlated with SAAM scores
at the moment. Participants receive the following instructions: [positively with SAAM security (r = .42, p < .05), and negatively with both SAAM
This scale consists of a number of adjectives that describe differ- anxiety and avoidance (r = .31, .30, respectively, both ps < .05)].
368 O. Gillath et al. / Journal of Research in Personality 43 (2009) 362373
lation between SAAM security and SAAM avoidance was negative Table 4
and moderate (r = .48, p < .05). There were no other signicant SAAM scale differences between participants selecting different RQ categories.
Table 3 8. Study 5
SAAM scale correlations with trait attachment measures: ECR dimensions and RQ
rated categories. Having established that the SAAM has discriminant and conver-
SAAM security SAAM anxiety SAAM avoidance gent validity, in Study 5 we tested the SAAMs ability to capture
uctuations in the sense of security and insecurity caused by lab-
ECR anxiety .32* .42* .26*
ECR avoidance .49* .03 .59* oratory manipulation (priming). To do so, we examined whether
RQ secure .54* .02 .37* activating participants attachment security schemas would result
RQ dismissive .07 .22* .28* in higher scores on the SAAM security scale. We hypothesized that
RQ preoccupied .27* .39* .13 participants exposed to a security prime would have higher SAAM
RQ fearful .38* .12 .45*
security score, and lower SAAM anxiety and avoidance scores than
*
p<.05. participants exposed to a neutral prime.
O. Gillath et al. / Journal of Research in Personality 43 (2009) 362373 369
also more secure than those who were exposed to other primes (all breakup from a close relationship partner, and describe the event
p < .05), F(3, 140) = 2.11, p < .05, Partial Eta Squared = .05. Finally, in detail. In the ticket priming condition (n = 46) participants were
while the tendency existed for SAAM anxiety scores to be lower asked to imagine or recall getting a speeding or parking ticket, and
in the security condition as well, the effect was not signicant, describe the event in detail. In the exam failure priming condition
F(3, 140) = .68, ns. (n = 46) participants were asked to recall or imagine a time when
Overall, the results of Study 6 provided further support for the they failed a test and describe it in detail. In the neutral control
feasibility of the SAAM to capture temporarily changes in attach- condition (n = 61) participants were asked to recall or imagine
ment working models. Moreover, Study 6 demonstrated that it is walking to school from their dormitory or apartment. Immediately
the exposure to a security prime, rather than to just a positive after the priming task participants completed the SAAM. Cronbach
mood, or a self-esteem prime, which leads participants to feel less alphas in this sample were adequate for the SAAM avoidance, anx-
avoidant and more secure (and to a limited extant less anxious). iety, and security scales (see Table 2 for alphas and inter-
Although both Study 5 and Study 6 demonstrated the ability to correlations.
shift temporarily the sense of security, they both did it to the same
direction (more secure or less anxious and avoidant). Study 7 was 10.2. Results and discussion
aimed to determine whether the SAAM can capture changes made
in the other direction and specically in levels of attachment To examine the effects of the priming manipulation on the
anxiety. SAAM scores we used a MANOVA, with the SAAM scores being
the dependent variables, and experimental condition being the
10. Study 7 independent variable. The MANOVA revealed a signicant main ef-
fect for prime type, F(9, 477) = 2.26, p < .05, Partial Eta
In Study 7, we extended our tests of the SAAMs ability to cap- Squared = .03. As shown in Table 8, in line with our hypotheses,
ture uctuations in attachment security and insecurity by examin- people who were exposed to the relationship breakup prime re-
ing whether activating insecure schemas would cause people to ported higher state attachment anxiety, F(3, 198) = 3.83, p < .05,
report higher levels of state anxiety and avoidance and lower levels Partial Eta Squared = .06, than people who were exposed to the
of security. Specically, we compared thoughts about separation neutral prime (p < .05). The two other priming conditions (failure
(an attachment-relevant negative event) with other, non-attach- and ticket) were not signicantly different from either the neutral
ment-related negative events (failing an exam or getting a parking or the breakup prime, suggesting that only the breakup priming
ticket). We hypothesized that participants exposed to the separa- condition affected SAAM anxiety scores. There were no other sig-
tion prime would have higher SAAM anxiety scores, and lower nicant effects.
SAAM security score than participants exposed to other control The results of Study 7 provided further support for the feasibil-
primes. Avoidant people seem to apply a unique meaning to the ity of the SAAM to capture temporary changes in attachment work-
word separation, and they devote extra resources to try and sup- ing models. Moreover, Study 7 demonstrated that it is possible to
press its effects on them (e.g., Edelstein & Gillath, 2008; Mikulin- capture decreases and not only increases in the temporary sense
cer, Gillath, & Shaver, 2002). We, therefore, left the effect of the of attachment security, and that attachment-related threat, but
separation prime on SAAM avoidance scores open. not non-attachment threat can lead to an increase in the SAAM
anxiety scores.
10.1. Method
11. General discussion
10.1.1. Participants
Two hundred and two undergraduates (139 women and 63 We reported seven studies describing the construction and val-
men) participated in the study for class credit. Their ages ranged idation of a new state scale the state adult attachment measure
from 18 to 32 (Mdn = 19). Fifty three percent were involved in a (SAAM). The SAAM lls the gap in the literature dealing with the
romantic relationship. Forty-three percent were Caucasian, 32% momentary level of the sense of attachment security and insecu-
Asian-American, 7% Latino, 7% African-American, and 11% did not rity, capturing uctuations in these dimensions as they occur in re-
specify their ethnicity. sponse to situational variables. The attachment literature seems to
be converging on the notion that each individual contains in his or
10.1.2. Materials and procedure her mind not one but several quite different working models that
In the experimental sessions participants were randomly as- can each be activated (or not) and applied (or not) as situations
signed to one of four conditions. In the separation priming condi-
tion (n = 49) participants were asked to imagine or recall a
Table 8
Effects of priming condition on SAAM scores in Study 7.
Table 7
Effects of priming condition on SAAM scores in Study 6. Measures Priming conditions
Breakup Failure Ticket Neutral (walking)
Measures Priming Conditions
SAAM avoidance
Security Self-esteem Positive mood Neutral (walking)
M 2.87 2.42 2.93 2.81
SAAM avoidance SD 1.22 .81 1.04 1.14
M 2.48 3.03 3.01 3.21
SAAM anxiety
SD 1.03 .69 .86 .90
M 4.46a 4.05b 4.08ab 3.70ab
SAAM anxiety SD 1.06 1.08 1.11 1.38
M 4.12 4.44 4.19 4.23
SAAM security
SD 1.06 1.15 .95 1.06
M 5.50 5.64 5.48 5.74
SAAM security SD .97 1.06 1.12 .93
M 5.68 5.18 5.23 5.20
SD .82 .99 .94 .76 Different letters mean this condition was different at a .05 signicance level than a
condition with a different letter.
O. Gillath et al. / Journal of Research in Personality 43 (2009) 362373 371
and relationships demand. The SAAM is compatible with this new- having a troubled quality to them. For instance, I really need
er conception of the exible nature of working models, and can as- someones emotional support and I wish someone would tell
sess their temporary experience. Throughout the studies the SAAM me they really love me. Interestingly, however, the items pertain-
factors showed good internal reliability, discriminant, convergent, ing to concern about others availability loaded onto the security
and criterion validity, and a moderate stability (as would be ex- factor, indicating that there is a difference between intense activa-
pected from a state measure). The SAAM was also sensitive to con- tion of the attachment system and concerns about whether attach-
ceptually relevant cues, with a distinct and theoretically coherent ment gures will respond to that activation.
pattern of results for each one of the factors. The fact that we did not also nd a fearful avoidant factor is in
line with the emotion literature. When measuring state affect
11.1. Two ways in which attachment changes over time researchers often nd (e.g., Russell & Carroll, 1999) that one cannot
be high on two different emotions in the same moment (in other
There are different ways of identifying change in attachment. words, one cannot be really happy and really anxious/upset simul-
One represents a momentary disequilibrium or shift in working taneously). Similarly, it seems that at least in the state level people
models, which is situation-specic and transitory. Another repre- do not feel both inclined and reluctant to get closer and intimate
sents trait-like change, a systematic reorganization of ones work- with their relationship partners at the same time. Rather it seems
ing models of self and other in general, which tends to be relatively to be one or the other.
persistent. An analogous differentiation between these two levels There are a few limitations to the studies described above. First,
of change can be seen in the study of personality traits. In his work all the scales are unipolar, that is, there are no reverse scored items
on personality states, Fleeson (2001) applied the Big Five trait tax- on any of them. While constructing the measure we did include
onomy to momentary behavior using experience-sampling meth- such reverse scored items (either taken from the ECR or from other
ods and found that peoples behaviors across a couple of weeks places), they did not make it into the nal version of the measure.
varied widely. However, when he examined peoples central ten- This might be due to the differences between the behavioral (i.e.,
dencies (means) of trait behaviors from the rst week to the sec- momentary) structure and the trait structure (although it should
ond week, he found stability coefcients around .90, which are be noted that even in the ECR there is only one reverse scored item
higher than most correlations in psychology. In other words, on the anxiety scale, perhaps suggesting that this is an underlying
whereas peoples momentary behaviors related to any one trait problem in attachment scale construction). A second limitation is
uctuated widely (and changed according to the situations they the differences in the correlations of security and avoidance, and
found themselves in; Fleeson, 2007), their mean levels of traits security and anxiety across studies. While the specic correlations
stayed pretty much the same. The SAAM scales assess the rst type were not the same, the overall structure and links between the
of change momentary uctuations in working models. scales were similar across the seven studies, creating a Y like shape
between the three scales.
11.1.1. Structure of the SAAM Finally our model might not map on precisely to the trait model
Various researchers have argued that the attachment construct as represented by the ECR. This will require measuring longitudinal
consists of three factors. For example, Simpson and his colleagues relations between the state and trait data. For instance, the SAAM
(e.g., Simpson, Rholes, & Nelligan, 1992) called their two dimensions should exhibit more short-term variance than the ECR, but over
Security versus Avoidance and Anxiety (about abandonment). time, average SAAM scores should converge meaningfully with
Collins and Read (1990) called their three dimensions Close, ECR scores. Clearly, future work should address this and provide
Depend, and Anxiety (about abandonment). Hazan and Shaver more details on the relation between states and traits. Future stud-
(1987), in their early model of adult attachment, used the terms that ies should also use methods such as experience-sampling studies
we use here: security, avoidance, and anxiety. Similar to these to chart how attachment changes over time, examine attachment
researchers, our analyses revealed a three-factor model, including situation-behavior contingencies (the inuence of momentary sit-
scales that assess temporary changes on anxiety, avoidance, and uational characteristics on state attachment levels), and use vari-
security, with the last two being moderately (negatively) correlated. ous contexts (such as therapy) to further capture the SAAM
Although the most recent measures of adult attachment style abilities and advantages. Future studies will also have to examine
(i.e., the ECR and the ECR-R) tend to have a two-dimensional struc- questions such as: Do those in a long-term relationship respond
ture, they measure traits or averaged attitudes and behaviors. At similarly to those in a short-term relationship or one that has re-
least based on the current studies, momentary changes to attach- cently ended? Did the primed subjects differ whether they were
ment take the form of a three dimensional construct. The SAAM actually in a good relationship versus and imagined relationship,
captures and differentiates between three unique psychological or had actually had an experience of breaking up?
processes: anxiety about attachment, avoidance of attachment, Despite some limitations, the results reported here are highly
and security-based strategies (Mikulincer & Shaver, 2003), the encouraging, supporting the reliability, validity and utility of the
latter of which is more than simply low anxiety and avoidance. SAAM as a state measure of attachment. Such a measure has not
As with other sensations, mood states, or emotions, these dimen- existed previously, and none of the existing attachment scales is
sions are somewhat independent, with unique patterns of associa- built in a way that allows measuring both generalized and momen-
tions with personality dimensions (e.g., Noftle & Shaver, 2006) and tary attachment. The new measure will allow researchers to assess
with approach and avoidance tendencies (Elliot & Reis, 2003). In whether various manipulations cause temporary changes to work-
other words, just like being low on trait negative affect does not ing models, allow clinicians to examine the efcacy of various
necessarily mean being high on trait positive affect; being low on treatments, and has the potential to advance the eld in the direc-
the anxiety and/or avoidance SAAM scales does not necessarily tion of understanding the dynamics of adult attachment.
mean being high on security. According to this reasoning, a state
measure of attachment requires measuring a third separate dimen- References
sion to represent security, which the SAAM does.
Another difference between the SAAM and trait measure lies in Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns of
attachment: A psychological study of the strange situation. Oxford, UK: Erlbaum.
the anxiety dimension. The SAAM anxiety items seem to reect (as Asendorpf, J. B., Banse, R., Wilpers, S., & Neyer, F. J. (1997). Relationship-specic
intended) hyperactivation of the attachment system the quasi- attachment scales for adults and their validation with network and diary
desperate need to be close to and loved by a relationship partner; procedure. Diagnostica, 43, 289313.
372 O. Gillath et al. / Journal of Research in Personality 43 (2009) 362373
Baldwin, M. W., & Fehr, B. (1995). On the instability of attachment style ratings. Gillath, O., & Shaver, P. R. (2007). Effects of attachment style and relationship
Personal Relationships, 2, 247261. context on selection among relational strategies. Journal of Research in
Baldwin, M. W., Fehr, B., Keedian, E., Seidel, M., & Thomson, D. W. (1993). An Personality, 41, 968976.
exploration of the relational schemata underlying attachment styles: Self- Gorsuch, R. L. (1983). Factor analysis (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum.
report and lexical decision approaches. Personality and Social Psychology Bulletin, Hammond, J. R., & Fletcher, G. J. O. (1991). Attachment styles and relationship
19, 746754. satisfaction in the development of close relationships. New Zealand Journal of
Baldwin, M. W., Keelan, J. P. R., Fehr, B., Enns, V., & Koh-Rangarajoo, E. (1996). Social- Psychology, 20, 5662.
cognitive conceptualization of attachment working models: Availability and Hart, J., Shaver, P. R., & Goldenberg, J. L. (2005). Attachment, self-esteem,
accessibility effects. Journal of Personality and Social Psychology, 71, 94109. worldviews, and terror management: Evidence for a tripartite security
Banse, R. (2004). Attachment style and marital satisfaction: Evidence for dyadic system. Journal of Personality and Social Psychology, 88, 9991013.
conguration effects. Journal of Social and Personal Relationships, 21, 273282. Hazan, C., & Shaver, P. R. (1987). Romantic love conceptualized as an attachment
Bartholomew, K., & Horowitz, L. M. (1991). Attachment styles among young adults: process. Journal of Personality and Social Psychology, 52, 511524.
A test of a four-category model. Journal of Personality and Social Psychology, 61, Heatherton, T. F., & Polivy, J. (1991). Development and validation of a scale for
226244. measuring state self-esteem. Journal of Personality and Social Psychology, 60,
Bowlby, J. (1969/1982). Attachment and loss. Attachment (Vol. 1). New York: Basic 895910.
Books (1st ed. pub. 1969; 2nd ed. pub. 1982). Hughes, D. A. (2007). Attachment-focused family therapy. New York: W.W. Norton.
Bowlby, J. (1973). Attachment and loss. Separation: Anxiety and anger (Vol. 2). New Johnson, S. M., & Whiffen, V. (2003). Attachment processes in couples and families.
York: Basic Books. New York: Guilford Press.
Bowlby, J. (1980). Attachment and loss. Loss: Sadness and depression (Vol. 3). New Jreskog, K. G., & Srbom, D. (1996). LISREL 8: Users reference guide. Chicago:
York: Basic Books. Scientic Software International.
Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-report measurement of adult Karney, B. R., & Bradbury, T. N. (1995). The longitudinal course of marital quality
attachment: An integrative overview. In J. A. Simpson & W. S. Rholes (Eds.), and stability: A review of theory, methods, and research. Psychological Bulletin,
Attachment theory and close relationships (pp. 4676). New York: Guilford Press. 118, 334.
Carver, C. S. (1997). Adult attachment and personality: Converging evidence and a Kirkpatrick, L. A., & Hazan, C. (1994). Attachment styles and close relationships: A
new measure. Personality and Social Psychology Bulletin, 23, 865883. four-year prospective study. Personal Relationships, 1, 123142.
Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by Klohnen, E. C., & Bera, S. J. (1998). Behavioral and experiential patterns of avoidantly
the multitrait-multimethod matrix. Psychological Bulletin, 56, 81105. and securely attached women across adulthood: A 30-year longitudinal
Cattell, R. B. (1966). The scree test for the number of factors. Multivariate Behavioral perspective. Journal of Personality and Social Psychology, 74, 211223.
Research, 1, 245276. Klohnen, E. C., Weller, J. A., Luo, S., & Choe, M. (2005). Organization and predictive
Cattell, R. B. (1978). The Scientic Use of Factor Analysis in Behavioral and Life Sciences. power of general and relationship-specic attachment models: One for all, and
New York: Plenum. all for one? Personality and Social Psychology Bulletin, 31, 16651682.
Collins, N. L., & Read, S. J. (1990). Adult attachment, working models, and Kobak, R. R., Cole, H. E., Ferenz-Gillies, R., Fleming, W. S., & Gamble, W. (1993).
relationship quality in dating couples. Journal of Personality and Social Attachment and emotion regulation during mother-teen problem-solving: A
Psychology, 58, 644663. control theory analysis. Child Development, 64, 231245.
Collins, N. L., & Read, S. J. (1994). Cognitive representations of attachment: The Levy, M. B., & Davis, K. E. (1988). Lovestyles and attachment styles compared: Their
structure and function of working models. In K. Bartholomew & D. Perlman relations to each other and to various relationship characteristics. Journal of
(Eds.), Attachment processes in adulthood (pp. 5392). London: Jessica Kingsley. Social and Personal Relationships, 5, 439471.
Cozzarelli, C., Karafa, J. A., Collins, N. L., & Tagler, M. J. (2003). Stability and change in Mikulincer, M., Gillath, O., & Shaver, P. R. (2002). Activation of the attachment
adult attachment styles: Associations with personal vulnerabilities, life events system in adulthood: Threat-related primes increase the accessibility of mental
and global construals of self and others. Journal of Social and Clinical Psychology, representations of attachment gures. Journal of Personality and Social
22, 315346. Psychology, 83, 881895.
Davila, J., Burge, D., & Hammen, C. (1997). Why does attachment style change? Mikulincer, M., Hirschberger, G., Nachmias, O., & Gillath, O. (2001). The affective
Journal of Personality and Social Psychology, 73, 826838. component of the secure base schema: Affective priming with representations
Davila, J., & Sargent, E. (2003). The meaning of life (events) predicts change in of attachment security. Journal of Personality and Social Psychology, 80, 305321.
attachment security. Personality and Social Psychology Bulletin, 29, 13831395. Mikulincer, M., & Shaver, P. R. (2003). The attachment behavioral system in
Downey, G., & Feldman, S. I. (1996). Implications of rejection sensitivity for intimate adulthood: Activation, psychodynamics, and interpersonal processes. In M. P.
relationships. Journal of Personality and Social Psychology, 70, 13271343. Zanna (Ed.). Advances in experimental social psychology (Vol. 35, pp. 53152).
Edelstein, R. S., & Gillath, O. (2008). Avoiding interference. Adult attachment and New York: Academic Press.
emotional processing biases. Personality and Social Psychology Bulletin, 34, Mikulincer, M., & Shaver, P. R. (2007a). Boosting attachment security to promote
171181. mental health, prosocial values, and inter-group tolerance. Psychological Inquiry,
Elizur, Y., & Mintzer, A. (2003). Gay males intimate relationship quality: The roles of 18, 139156.
attachment security, gay identity, social support, and income. Personal Mikulincer, M., & Shaver, P. R. (2007b). Attachment patterns in adulthood: Structure,
Relationships, 10, 411435. dynamics, and change. New York: Guilford Press.
Elliot, A. J., & Reis, H. T. (2003). Attachment and exploration in adulthood. Journal of Mikulincer, M., Shaver, P. R., Gillath, O., & Nitzberg, R. E. (2005). Attachment,
Personality and Social Psychology, 85, 317331. caregiving, and altruism: Boosting attachment security increases compassion
Feeney, J. A., & Noller, P. (1992). Attachment style and romantic love: Relationship and helping. Journal of Personality and Social Psychology, 89, 817839.
dissolution. Australian Journal of Psychology, 44, 6974. Muthen, L. K., & Muthen, B. O. (2006). Mplus Users Guide (4th ed.). Los Angeles, CA:
Fleeson, W. (2001). Towards a structure- and process-integrated view of Muthen & Muthen.
personality: Traits as density distributions of states. Journal of Personality and Noftle, E. E., & Shaver, P. R. (2006). Attachment dimensions and the Big Five
Social Psychology, 80, 10111027. personality traits: Associations and comparative ability to predict relationship
Fleeson, W. (2007). Situation-based contingencies underlying trait-content quality. Journal of Research in Personality, 40, 179208.
manifestation in behavior. Journal of Personality, 75, 825862. Oppenheim, D., & Goldsmith, D. F. (Eds.). (2007). Attachment theory in clinical work
Fletcher, G. J. O., Simpson, J. A., & Thomas, G. (2000). The measurement of perceived with children: Bridging the gap between research and practice. New York: Guilford.
relationship quality components: A conrmatory factor-analytic approach. Overall, N. C., Fletcher, G. J. O., & Friesen, M. (2003). Mapping the intimate
Personality and Social Psychology Bulletin, 26, 340354. relationship mind: Comparisons between three models of attachment
Fraley, R. C. (2002). Attachment stability from infancy to adulthood: Meta-analysis representations. Personality and Social Psychology Bulletin, 29, 14791493.
and dynamic modeling of developmental mechanisms. Personality and Social Pierce, T., & Lydon, J. E. (2001). Global and specic relational models in the
Psychology Review, 6, 123151. experience of social interactions. Journal of Personality and Social Psychology, 80,
Fraley, R. C., & Brumbaugh, C. C. (2004). A dynamical systems approach to 613631.
understanding stability and change in attachment security. In W. S. Rholes & J. Rempel, J. K., & Holmes, J. G. (1986). How do I trust thee? Psychology Today, 20,
A. Simpson (Eds.), Adult attachment: Theory, research, and clinical implications 2834.
(pp. 86132). New York: Guilford Press. Rempel, J. K., Ross, M., & Holmes, J. G. (2001). Trust and communicated
Fraley, R. C., Waller, N. G., & Brennan, K. A. (2000). An item-response theory analysis attributions in close relationships. Journal of Personality and Social Psychology,
of self-report measures of adult attachment. Journal of Personality and Social 81, 5764.
Psychology, 78, 350365. Robins, R. W., Caspi, A., & Moftt, T. E. (2000). Two personalities, one relationship:
Fyffe, C., & Waters, E. (1997). Empirical classication of adult attachment status: Both partners personality traits shape the quality of their relationship. Journal
Predicting group membership. Poster presented at the biennial meeting of the of Personality and Social Psychology, 79, 251259.
Society for Research in Child Development, Washington, DC. Rothbart, M. K., Ahadi, S. A., & Evans, D. E. (2000). Temperament and personality:
George, C., Kaplan, M., & Main, M. (1985). Adult attachment interview. Unpublished Origins and outcomes. Journal of Personality and Social Psychology, 78, 122135.
manuscript, University of California, Berkeley. Russell, J. A., & Carroll, J. M. (1999). On the bipolarity of positive and negative affect.
Gillath, O., Mikulincer, M., Fitzsimons, G. M., Shaver, P. R., Schachner, D. A., & Bargh, Psychological Bulletin, 125, 330.
J. A. (2006). Automatic activation of attachment-related goals. Personality and SAS Institute, Inc. (20022004). SAS/STAT Users Guide, SAS OnlineDoc 9.1.3, Cary,
Social Psychology Bulletin, 32, 13751388. NC: SAS Institute, Inc.
Gillath, O., Selcuk, E., & Shaver, P. R. (2008). Moving toward a secure attachment Shaver, P. R., & Brennan, K. A. (1992). Attachment style and the big ve personality
style: Can repeated security priming help? Social and Personality Compass. traits: Their connection with romantic relationship outcomes. Personality and
doi:10.1111/j.1751-9004.2008.00120. Social Psychology Bulletin, 18, 536545.
O. Gillath et al. / Journal of Research in Personality 43 (2009) 362373 373
Simpson, J. A. (1990). Inuence of attachment styles on romantic relationships. of Grays anxiety and impulsivity dimensions. Personality and Individual
Journal of Personality and Social Psychology, 59, 971980. Differences, 31, 837862.
Simpson, J. A., Collins, W. A., Tran, S., & Haydon, K. C. (2007). Attachment and the Waters, E., Weineld, N. S., & Hamilton, C. E. (2000). The stability of attachment
experience and expression of emotions in adult romantic relationships: A security from infancy to adolescence and early adulthood: General discussion.
developmental perspective. Journal of Personality and Social Psychology, 92, Child Development, 71, 703706.
355367. Watson, D. (2004). Stability versus change, dependability versus error: Issues in the
Simpson, J. A., Rholes, W. S., & Nelligan, J. S. (1992). Support seeking and support assessment of personality over time. Journal of Research in Personality, 38,
giving within couples in an anxiety-provoking situation: The role of attachment 319350.
styles. Journal of Personality and Social Psychology, 62, 434446. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief
Spielberger, C. D. (1983). Manual for the state-trait anxiety inventory (STAI). Palo Alto, measures of positive and negative affect: The PANAS scales. Journal of
CA: Consulting Psychologists Press. Personality and Social Psychology, 54, 10631070.
Torrubia, R., vila, C., Molt, J., & Caseras, X. (2001). The sensitivity to
punishment and sensitivity to reward questionnaire (SPSRQ) as a measure