Relations Between Anxiety Sensitivity and Attachment Style Dimensions

Abstract

This study replicated and extended the work of C. F. Weems, S. L. Berman, W. K. Silverman, and E. T. Rodriquez (2002) by investigating relations between anxiety sensitivity (AS) and attachment dimensions in a sample of young adults. Two hundred and twenty-six undergraduate students completed self-report measures including the Anxiety Sensitivity Index and the measure of adult romantic attachment used by C. F. Weems et al. (2002). In order to investigate the association between AS and a different domain of attachment, a measure of adult attachment referring to close relationships was included. As defined by both measures, insecurely attached individuals, specifically those classified as preoccupied and fearful (i.e., those with negative Models of Self), reported significantly higher levels of AS than those with secure and dismissing attachment styles (i.e., those with positive Models of Self). Results indicated that across both measures the Model of Self attachment dimension accounted for unique variance in AS levels beyond that contributed by trait anxiety. The Model of Others attachment dimension had a more limited association with AS.

Full text

P1: IZO
joba2004.cls (03/17/2004 v1.1 LaTeX2e JOBA document class) pp1397-joba-NY00000635 October 9, 2004 20:22
FOR PROOFREADING ONLY
Journal of Psychopathology and Behavioral Assessment, Vol. 27, No. 3, September 2005 ( C
2005)
Relations Between Anxiety Sensitivity and Attachment
Style Dimensions
Margo C. Watt,1,2Lachlan A. McWilliams,1and Anna G. Campbell1
Accepted
This study replicated and extended the work of C. F. Weems, S. L. Berman, W. K. Silverman, and
E. T. Rodriquez (2002) by investigating relations between anxiety sensitivity (AS) and attachment
dimensions in a sample of young adults. Two hundred and twenty-six undergraduate students com-
pleted self-report measures including the Anxiety Sensitivity Index and the measure of adult romantic
attachment used by C. F. Weems et al. (2002). In order to investigate the association between AS and
a different domain of attachment, a measure of adult attachment referring to “close relationships” was
included. As defined by both measures, insecurely attached individuals, specifically those classified as
preoccupied and fearful (i.e., those with negative Models of Self), reported significantly higher levels
of AS than those with secure and dismissing attachment styles (i.e., those with positive Models of
Self). Results indicated that across both measures the Model of Self attachment dimension accounted
for unique variance in AS levels beyond that contributed by trait anxiety. The Model of Others
attachment dimension had a more limited association with AS.
KEY WORDS: anxiety; sensitivity; attachment; Model of Self; Model of Others.
Anxiety sensitivity (AS) is an individual difference
P1
variable characterized by a fear of anxiety-related sensa-
tions arising from beliefs that these sensations have harm-
ful consequences (Reiss, 1991). AS is conceptualized
as being composed of three intercorrelated lower-order
components that load on a common higher-order factor
(i.e., the global AS construct; see Zinbarg, Mohlman, &
Hong, 1999). These three lower-order components are
commonly referred to as Physical Concerns (e.g., fears
of anxiety-related physical sensations such as a racing
heart), Psychological Concerns (e.g., fears of cognitive
symptoms of anxiety such as difficulty concentrating),
and Social Concerns (e.g., fears of publicly observable
anxiety symptoms such as shaking). AS has been found to
play a role in the development of panic attacks (Schmidt,
Lerew, & Jackson, 1997) and to be positively associated
with other anxiety and mood disorders (see Cox, Borger,
& Enns, 1999).
1Department of Psychology, St. Francis Xavier University, Antigonish,
NS, Canada.
2To whom correspondence should be addressed at Department of
Psychology, St. Francis Xavier University, P.O. Box 5000, Antigonish,
NS, Canada B2G 2W5; e-mail: mwatt@stfx.ca.
Research regarding the development of AS has fo-
cused on genetic factors (Stein, Jang, & Livesley, 1999),
specific types of learning experiences (Watt, Stewart,
& Cox, 1998), and attachment styles (Weems, Berman,
Silverman, & Rodriquez, 2002). This study further inves-
tigated the relationships between attachment character-
istics and AS. Attachment theory (Bowlby, 1988) posits
that primary caregivers’ patterns of responsiveness to their
children establish enduring expectancies regarding how
others will respond when one is in distress. Consistent
and responsive care is thought to foster secure attachment
(i.e., a Model of Self as valued and self-sufficient as well
as a Model of Others as caring and trustworthy); whereas
less responsive and/or rejecting behavior from caregivers
is thought to result in insecure attachment (i.e., negative
internalized models of self and others). A large litera-
ture has extended attachment theory to adult relationships
(e.g., Hazan & Shaver, 1987).
The most commonly used method of assessing
adult attachment styles are self-report measures based on
Bartholomew and Horowitz’s (1991) conceptualization of
adult attachment as being composed of two related dimen-
sions referring to one’s internal Model of Self and one’s
internal Model of Others. The Model of Self dimension
191
0882-2689/05/0900-0191/0 C
2005 Springer Science+Business Media, Inc.

P1: IZO
joba2004.cls (03/17/2004 v1.1 LaTeX2e JOBA document class) pp1397-joba-NY00000635 October 9, 2004 20:22
192 Watt, McWilliams, and Campbell
refers to a continuum ranging from no anxiety about re-
jection or abandonment to intense anxiety about rejection
based on beliefs of personal unworthiness. The Model
of Others dimension refers to a continuum ranging from
interpersonal trust to mistrust, avoidance of others, and
discomfort with interpersonal closeness. Each dimension
can be dichotomized as positive or negative yielding four
possible combinations referred to as attachment styles.
These styles include secure (positive self and positive
others), preoccupied (negative self and positive others),
fearful (negative self and negative others), and dismiss-
ing (positive self and negative others) attachment styles.
Although secure attachment is considered to be a pro-
tective factor, insecure attachment has been found to be
associated with psychopathology (Mickelson, Kessler, &
Shaver, 1997; Rutter, 1997).
Silverman and Weems (1999) first raised the possi-
bility that insecurely attached individuals could be pre-
disposed to misinterpret benign symptoms of anxiety as
being catastrophic. On the basis of this suggestion, Weems
et al. (2002) used two student samples (viz., high school
and undergraduate students) to test the hypothesis that
insecurely attached individuals have higher levels of AS.
Those classified as preoccupied and fearful scored sig-
nificantly higher on the Anxiety Sensitivity Index (ASI;
Peterson & Reiss, 1992) than those with secure and dis-
missing attachment styles. Model of Self was significantly
and positively correlated with ASI total and subscale
scores (rs ranged from .27 to .41). Partial correlations
indicated that Model of Self scores could account for
significant variance in ASI total scores and several ASI
subscales after statistically controlling for current anxi-
ety symptoms and general psychological distress as mea-
sured by the Symptom Checklist-90 (SCL-90; Derogatis,
1983). Correlations and partial correlations indicated that
the Model of Others had weak and largely nonsignificant
associations with the AS variables.
Three issues not considered by Weems et al. (2002)
were the focus of this study. First, Weems et al. used a
measure of romantic attachment. Although this was not
a limitation, it is important to determine whether AS is
similarly associated with nonromantic forms of attach-
ment. For example, Weems et al. did not find a strong
association between Model of Others and AS, but it re-
mained possible that attachment beliefs regarding close
relationships in general might be more salient than ro-
mantic attachment beliefs when considering expectancies
regarding the amount of assistance others might provide
during an episode of anxiety. In order to investigate this
hypothesis, a measure of adult attachment referring to
“close relationships” was included. This hypothesis would
be supported if the Model of Others dimension of the
general attachment scale was more strongly correlated
with AS than the Model of Others dimension of the ro-
mantic attachment measure.
Second, Weems et al. (2002) suggested that inse-
curely attached individuals may distort, selectively en-
code, and be more fearful of anxiety-related sensations
because they think they are unable to help themselves or
obtain help from others. However, it is possible that these
anxious and negative perceptions may be due to a more
general tendency to experience anxiety or negative af-
fect (e.g., trait anxiety) and that the association observed
between attachment and AS are the result of common
variance shared among all three constructs. Viewed from
this perspective, it was important to determine whether
attachment measures could account for unique variance
in ASI scores beyond that accounted for by trait anxiety.
Although Weems et al. employed the SCL-90 to statis-
tically control for current anxiety symptoms and general
psychological distress, this study employed the State Trait
Anxiety Inventory—Trait (STAI-T; Spielberger, Gorsuch,
Lushene, Vagg, & Jacobs, 1983). Trait anxiety, as mea-
sured by the STAI-T, was selected because it is the most
commonly used construct when assessing the incremental
validity of AS in predicting psychopathology and poten-
tial interaction effects.
Third, Weems et al. (2002) stated that their findings
were consistent with the idea that a negative Model of
Others could lead to a belief that anxiety symptoms will re-
sult in “social censure” and that “ ...interpersonal distrust
may heighten anxiety symptom related distress because
an individual may feel they may not get help from others”
(p. 166). Their correlation analyses did not provide strong
support for this idea, but their multiple regression analyses
revealed that the interaction between anxiety symptoms
and Model of Others was a significant positive predictor
of ASI total scores. This finding raised the possibility
that, among those with elevated levels of anxiety, Model
of Others would be associated with AS. This issue was
investigated further in this study by determining whether
this interaction effect could be replicated and whether it
applied to each of the lower-order components of AS.
METHOD
Participants
Participants were 226 students (76 men, 150 women),
enrolled in introductory psychology courses at an eastern
Canadian university. Participants were predominantly of
Euro Canadian descent (98%) and had a mean age of
19.3 years (range: 18–29 years of age).

P1: IZO
joba2004.cls (03/17/2004 v1.1 LaTeX2e JOBA document class) pp1397-joba-NY00000635 October 9, 2004 20:22
Anxiety Sensitivity and Attachment 193
Measures
Experiences in Close Relationships (ECR; Brennan,
Clark, & Shaver, 1998)
The ECR includes two 18-item scales measuring the
two dimensions underlying adult attachment styles. Re-
spondents are asked to indicate their level of agreement
with a series of statements concerning their relationship
experiences using a response format ranging from 1 (dis-
agree strongly)to7(agree strongly). Continuous mea-
sures of Model of Others and Model of Self are created by
averaging the scores of the items from these two scales.
Respectively, Brennan et al. labeled the Model of Self
and Model of Others scales as the Anxiety and Avoidance
scales (i.e., their affective-behavioral names). In order to
maintain consistency with terminology used by Weems
et al. (2002), the labels Model of Self and Model of Oth-
ers were employed in this study with higher scores on
these scales indicating more negative models. Brennan
et al.’s (1998) method of assigning individuals to one of
the attachment categories based on their scores on the two
continuous measures (e.g., those with positive Models of
Self and Others are classified as securely attached) was
also used.
Relationship Questionnaire (RQ; Bartholomew &
Horowitz, 1991)
The RQ consists of four short paragraphs each de-
scribing a prototypical attachment pattern (i.e., secure,
fearful, preoccupied, and dismissing attachment styles)
regarding close relationships in general. Participants rate
the degree to which each of the paragraphs describes
their general relationship styles. These ratings provide
continuous scores of each of the four attachment pat-
terns and can be used to categorize participants into their
best fitting attachment pattern according to their highest
rated attachment description. Griffin and Bartholomew
(1994) validated the four RQ attachment statements using
semistructured attachment interviews and concluded that
the two underlying dimensions could be reliably assessed
by self-report. They also found the RQ to demonstrate
high predictive validity.
Relationship Scales Questionnaire (RSQ; Griffin &
Bartholomew, 1994)
The RSQ contains 30 short statements that ask partic-
ipants to rate, on a 5-point scale, the extent to which they
believe each statement best describes their feelings about
close relationships. The totals of the four or five items
that represent each attachment prototype are calculated in
order to determine scores for each attachment pattern.
For the purposes of this study, the RQ and the RSQ
were combined to obtain a composite measure of adult
attachment as recommended by Bartholomew and Shaver
(1998). Attachment ratings on both the RQ and the RSQ
were converted into standard scores (zscores), and then
the standardized parallel RQ and RSQ scores were com-
bined to form single composite scores. The composite
attachment ratings were used to obtain the Model of Self
and Model of Others attachment dimensions. Unlike the
ECR, higher scores on the RQ-RSQ dimensions repre-
sented more positive Models of Self and Others. Indi-
viduals with high Self and Others scores were classified
as securely attached, individuals with high Self and low
Others scores were classified as dismissing, individuals
with low Self and high Others scores were classified as
preoccupied, and individuals with low Self and Others
scores were classified as fearful.
Anxiety Sensitivity Index (ASI; Peterson & Reiss, 1992)
The ASI is a 16-item measure of fear of anxiety-
related signs and symptoms. Participants rate the extent
to which they agree or disagree with each item pertain-
ing to beliefs that anxiety symptoms are signs of harm-
ful/aversive consequences by selecting one of five points
on a Likert scale that ranges from 0 (very little)to4
(very much). Considerable evidence indicates that ASI has
strong psychometric properties (see reviews by Peterson
& Reiss, 1992; Taylor, 1995). ASI subscales representing
the lower-order components of AS were created using the
scoring system recommended by Zinbarg et al. (1999).
The Physical Concerns subscale included eight items
(items 3, 4, 6, 8, 9, 10, 11, and 14), the Psychological Con-
cerns subscale included four items (items 2, 12, 15, and
16), and the Social Concerns subscale included four items
(items 1, 5, 7, and 13). Respectively, the internal consisten-
cies (Cronbach’s alphas) of the Physical, Psychological,
and Social Concerns subscales in this study were .85,
.76, and .55. Given the differing number of items, ASI
subscales were scored as item means (possible range:
0–4).
State-Trait Anxiety Inventory—Trait Scale (STAI-T)
The STAI-T (Spielberger et al., 1983) is a 20-
item self-report measure on which participants rate how

P1: IZO
joba2004.cls (03/17/2004 v1.1 LaTeX2e JOBA document class) pp1397-joba-NY00000635 October 9, 2004 20:22
194 Watt, McWilliams, and Campbell
anxious they “generally” feel on 4-point Likert-type scales
of relative frequency that range from 1 (almost never)to4
(almost always). Considerable evidence indicates that the
STAI-T has strong psychometric properties (see review by
Spielberger et al., 1983). The STAI-T was included in this
study as a general measure of the tendency to experience
anxiety or negative affect in stressful circumstances.
Procedure
This study was given ethics clearance by the Depart-
ment of Psychology’s Research Ethics Committee. Partic-
ipants were first-year psychology students recruited from
the Psychology Department’s subject pool who received
course credit in their introductory psychology course as
compensation for their participation. The general purpose
of the study was explained to potential participants. Those
agreeing to participate provided written informed consent
and were assured that they were free to decline participa-
tion at any time prior to completion of the study without
loss of the expected compensation. The questionnaires
were administered outside of class time to small groups
of 5–20 students.
RESULTS
Anxiety Characteristics
Means and standard deviations of variables of inter-
est (age, ASI, and STAI-T scores) are presented in Table I.
Women scored significantly higher than did men on both
the Total ASI [Mean =19.90 vs. 15.92, F(1,226) =9.37,
p<.01] and ASI-Physical Concerns scores [Mean =.70
vs. .65, F(1,226) =9.37, p<.01]. There was no sig-
nificant difference between men and women in reporting
of ASI-Psychological and ASI-Social concerns [Mean =
1.27 vs. 0.88, F(1,226) =0.27, ns; and 1.73 vs. 1.66,
F(1,226) =0.46, ns, respectively]. These results are sim-
ilar to the findings of other studies using samples of uni-
versity students (e.g., Watt & Stewart, 2003)
Comparison of Attachment Measures
Using ECR scores, participants were most frequently
classified as having a secure attachment style (39%).
The remaining participants were divided approximately
equally across the three insecure attachment dimensions
(see Table I). Similar percentages were found using com-
bined RQ-RSQ scores and are also reported in Table I.
Chi-square analysis indicated that the classifications ob-
tained from the two attachment measures were signifi-
cantly related [χ2(7) =147.91, p<.001]. Of the people
who were classified as secure on the ECR, 70% were clas-
sified as secure on the RQ-RSQ. Of the people who were
classified as fearful on the ECR, 46% were classified as
fearful and 36% were classified as preoccupied on the RQ-
RSQ. Of the people who were classified as preoccupied
on the ECR, 49% were classified as preoccupied and 35%
were classified as fearful on the RQ-RSQ. Of the people
Tab l e I . Descriptive and ANOVA Results for Both Attachment Measures
N%MF Age ASI STAI-T
Secure
ECR 88 39 23 65 19.0a(1.3) 14.7 (7.9)a,b34.4 (7.7)a
RQ-RSQ 72 37 20 52 19.2 (1.3) 15.4 (8.7)a35.5 (8.3)a
Fearful
ECR 46 20 20 26 20.0a(2.5) 23.6 (9.0)a,c46.4 (10.1)a,c
RQ-RSQ 38 20 11 37 19.6 (2.1) 19.4 (10.2) 44.1 (8.5)a,c
Preoccupied
ECR 52 23 12 40 19.2 (2.1) 22.4 (9.1)b,d45.4 (8.9)b,d
RQ-RSQ 46 23 14 32 19.5 (2.7) 22.8 (8.1)a,b46.3 (9.3)b,d
Dismissing
ECR 39 18 21∗18 19.4 (1.3) 15.5 (6.9)c,d38.1 (7.5)c,d
RQ-RSQ 40 20 21∗19 19.5 (2.0) 17.2 (7.5)b37.4 (9.3)c,d
Total
ECR 225 100 76 149 19.4 (1.8) 18.4 (9.1) 40.0 (9.8)
RQ-RSQ 196 100 66 130 19.4 (1.9) 18.3 (9.2) 40.0 (9.9)
Note.ECR=Experiences in Close Relationships Inventory; RQ-RSQ =Relationship Questionnaire-
Relationship Scales Questionnaire. Identical subscripts within columns indicate significant differ-
ences on Tukey’s contrasts.
∗Indicates significant difference between males and females (p<.05).

P1: IZO
joba2004.cls (03/17/2004 v1.1 LaTeX2e JOBA document class) pp1397-joba-NY00000635 October 9, 2004 20:22
Anxiety Sensitivity and Attachment 195
who were classified as dismissing on the ECR, 59% were
classified as dismissing and 22% were classified as secure
on the RQ-RSQ.
As expected, the ECR Model of Self scale had a
strong negative correlation with the corresponding scale
from the RQ-RSQ [r=−.75, p<.001] and the ECR
Model of Others scale had a strong negative correla-
tion with the corresponding RQ-RSQ scale [r=−.58,
p<.001]. The ECR Model of Self scale had a signif-
icant positive association with the complementary ECR
Model of Others scale [r=.19, p<.01] and a non-
significant association with the RQ-RSQ Model of Others
scale [r=−.09, ns]. The ECR Model of Others scale
had a significant positive association with the RQ-RSQ
Model of Self scale [r=.28, p<.001]. This pattern of
findings indicates that the corresponding scales from the
two measures share common variance and that the two
attachment dimensions (Model of Self and Others) are
relatively distinct from each other.
Anxiety Measures and Attachment Categories
A one-way ANOVA revealed a significant differ-
ence in age across the four ECR attachment categories
[F(3,221) =3.48, p<.05]. Post hoc analyses utiliz-
ing the Tukey–Kramer modification of Tukey’s Honestly
Significant Difference Test (Tukey HSD) procedure in-
dicated that participants classified as fearful were sig-
nificantly older than participants classified as secure.
There was a significant effect of attachment category on
ASI scores [F(3,224) =17.88, p<.001] and STAI-T
scores [F(3,219) =28.71, p<.001]. Descriptive statis-
tics across attachment categories are presented in Table I.
Post hoc analyses indicated that participants who were
classified as fearful or preoccupied scored significantly
higher on both the ASI and STAI-T than participants
classified as either securely attached or dismissing.Chi-
square analyses revealed a significant difference between
attachment categories in terms of gender [χ2(3) =8.64,
p<.05]. An examination of standardized residuals indi-
cated that men were disproportionately represented in the
dismissing category (std. res. =2.2).
Similar analyses with the RQ-RSQ scores revealed
no significant difference in age across the four attachment
categories [F(3,193) =0.44, ns], but significant effects
of attachment category were found for ASI [F(3,195) =
6.99, p<.001] and STAI-T scores [F(3,190) =17.72,
p<.001]. Participants classified as preoccupied scored
significantly higher on the ASI than participants classi-
fied as either secure or dismissing. Participants classified
as fearful or preoccupied scored significantly higher on
the STAI-T than participants classified as either secure or
dismissing. Chi-square analyses revealed a significant dif-
ference between attachment categories in terms of gender
[χ2(3) =8.06, p<.05] with men being disproportion-
ately represented in the dismissing category (std. res. =
2.1; see Table I).
Anxiety Measures and Attachment Dimensions
Correlation coefficients were calculated to examine
the association between AS and the continuous attachment
dimensions assessed by the ECR and RQ-RSQ. These
findings are presented in Tables II and III, respectively,
and indicated that for both attachment measures, Model
of Self was consistently associated with the AS vari-
ables. Across both attachment measures, Model of Others
had more limited associations with the AS variables. To
further explore relations between AS and attachment, a
measure of trait anxiety was included in order to deter-
mine whether attachment dimensions could account for
unique variance in AS beyond that contributed by the gen-
eral tendency to experience anxiety. Partial correlations
Table II. Zero-Order and Partial (Controlling for STAI-T) Correlations Between ASI Scores
and ECR Attachment Dimensions
12 3 456
1. Model of Self — .06 .29∗∗∗ .18∗∗ .21∗∗ .34∗∗∗
2. Model of Others .20∗∗ —−.04 .24∗∗∗ .06 .07
3. ASI-Physical .40∗∗∗ .05 — .41∗∗∗ .40∗∗∗ .91∗∗∗
4. ASI-Psychological .43∗∗∗ .35∗∗∗ .50∗∗∗ —.33
∗∗∗ .64∗∗∗
5. ASI-Social .27∗∗∗ .11 .44∗∗∗ .37∗∗∗ —.59
∗∗∗
6. ASI-Total .49∗∗∗ .19∗∗ .92∗∗∗ .72∗∗∗ .60∗∗∗ —
Note. STAI-T =State Trait Anxiety Inventory—Trait Scale; ASI =Anxiety Sensitivity
Index; ECR =Experiences in Close Relationships Inventory. Partial correlations controlling
for STAI-T above the diagonal.
∗p<.05. ∗∗ p<.01. ∗∗∗p<.001.

P1: IZO
joba2004.cls (03/17/2004 v1.1 LaTeX2e JOBA document class) pp1397-joba-NY00000635 October 9, 2004 20:22
196 Watt, McWilliams, and Campbell
Table III. Zero-Order and Partial (Controlling for STAI-T) Correlations Between ASI Scores
and RQ-RSQ Attachment Dimensions
123456
1. Model of Self — .06 −.17∗∗ −.11 −.16∗∗ −.20∗∗
2. Model of Others .19∗∗ —.13
∗−.02 .02 .07
3. AS-Physical −.32∗∗∗ .03 — .41∗∗∗ .40∗∗∗ .92∗∗∗
4. AS-Psychological −.41∗∗∗ −.17∗.55∗∗∗ —.32
∗∗∗ .64∗∗∗
5. AS-Social −.24∗∗∗ −.03 .44∗∗∗ .37∗∗∗ —.58
∗∗∗
6. ASI-Total −.40∗∗∗ −.06 .92∗∗∗ .76∗∗∗ .60∗∗∗ —
Note. STAI-T =State Trait Anxiety Inventory—Trait Scale; ASI =Anxiety Sensitivity Index;
RQ-RSQ =Relationship Questionnaire–Relationship Scales Questionnaire. Partial correlations
controlling for STAI-T above the diagonal.
∗p<.05. ∗∗ p<.01. ∗∗∗p<.001.
(statistically controlling for STAI-T) between ASI scores
and the continuous attachment dimensions are also pre-
sented in Tables II and III. For both measures, partial
correlations indicated that the attachment dimensions did
account for significant variance in ASI total and subscale
scores after statistically controlling for STAI-T. In partic-
ular, the Model of Self dimension remained significantly
associated with ASI Total, Physical, Psychological, and
Social Concerns subscales after controlling for trait anxi-
ety. However, with the ECR only one of the partial correla-
tions involving the Model of Others dimension remained
significant with about 6% of the variance in Psychological
Concerns scores accounted for by scores on the Model of
Others dimension. Similarly, with the RQ-RSQ, only one
of the partial correlations involving the Model of Oth-
ers dimension remained significant with about 2% of the
variance in Physical Concerns scores accounted for by the
Model of Others dimension.
Hierarchical Regression Analyses With
Attachment Dimensions
Hierarchical multiple linear regression analyses were
conducted to investigate whether attachment beliefs re-
garding self and others could account for unique vari-
ance in the prediction of ASI total and subscale scores
beyond that contributed by the demographic variables
(age, gender) and trait anxiety (STAI-T), and to examine
whether the interaction between the attachment dimen-
sions and trait anxiety were associated with the AS vari-
ables. This approach represented a replication of Weems
et al.’s (2002) analyses, as well as an extension by in-
cluding ASI subscales as criterion variables. In all cases,
demographic variables and STAI-T scores were entered
in Steps 1 and 2, respectively, followed by a series of
separate steps which included one of the attachment di-
mensions (Step 3) followed by its interaction with STAI-T
scores (Step 4).3These regression models are summarized
in Tables IV to VII.
With regard to the ECR measures, the inclusion of
the Model of Self variable significantly improved all of
the regression models. The R2-change values indicated
that the inclusion of the Model of Self variable accounted
for an additional 9% of variance in ASI total scores and
also accounted for additional variance in AS-Physical,
-Psychological, and -Social Concerns subscales (respec-
tively, 7, 3, and 4%). In the case of the RQ-RSQ measures,
the inclusion of the Model of Self variable significantly
improved the regression models for each criterion except
AS-Psychological Concerns. The R2-change values indi-
cated that the inclusion of the Model of Self variable ac-
counted for an additional 3% variance in ASI total scores
and 3% of the additional variance in both the AS-Physical
and -Social Concerns subscales, respectively. The ECR
Model of Others dimension was examined in a similar
manner and was found to have a significant positive asso-
ciation with AS-Psychological Concerns only, accounting
for 3% of additional variance. The RQ-RSQ Model of
Others dimension was not significantly associated with
ASI total or subscale scores. Across both measures, the
interaction terms were not significantly associated with
the any of the criterion variables.
DISCUSSION
This study replicated the findings of Weems et al.
(2002) and extended their findings with the inclusion of
3Typically, interaction terms are highly correlated with their constituents,
a situation that can produce multicollinearity and unstable regression
estimates unless the terms have been centered. Centering refers to the
conversion of deviation scores so that each variable has a mean of zero
(Aiken & West, 1991). The Aiken and West approach of centering (or
“de-meaning”) involves subtracting the mean of each variable from the
values for each observation. This approach was used in this study and
produced results identical to original analyses.

P1: IZO
joba2004.cls (03/17/2004 v1.1 LaTeX2e JOBA document class) pp1397-joba-NY00000635 October 9, 2004 20:22
Anxiety Sensitivity and Attachment 197
Tab l e I V. Summary of Hierarchical Regression Analyses Predicting ASI Total Scores
Step Model R2Change in R2Fp<B
1. Age .00 .00 0.05 ns .02
1. Gender .04 .04 8.93 .01 .21
2. STAI-T .21 .17 44.53 .001 .41
3. ECR Model of Self .24 .09 26.94 .001 .35
4. ECR Model of Self ×STAI-T .24 .00 0.73 ns −.29
3. ECR Model of Others .22 .01 1.12 ns .07
4. ECR Model of Others ×STAI-T .22 .00 0.05 ns .00
3. RQ-RSQ Model of Self .30 .03 7.99 .01 −.21
4. RQ-RSQ Model of Self ×STAI-T .30 .00 0.35 ns .15
3. RQ-RSQ Model of Others .22 .00 1.09 ns .07
4. RQ-RSQ Model of Others ×STAI-T .22 .00 0.01 ns .02
Note. ASI =Anxiety Sensitivity Index; STAI-T =State Trait Anxiety Inventory—Trait Scale; ECR =
Experiences in Close Relationships Inventory; RQ-RSQ =Combined zscores for the Relationship
Questionnaire and Relationship Scales Questionnaire.
a measure of adult attachment referring to close relation-
ships in general (i.e., RQ-RSQ). Categorical analyses re-
vealed that most respondents were classified similarly on
both attachment measures with the greatest convergence
found with the secure and dismissing categories. For ex-
ample, 70% of individuals classified as secure with the
ECR were classified as secure with the RQ-RSQ. Corre-
lations between the scales of each measure revealed that
the corresponding scales from the two measures shared
common variance and that the two attachment dimen-
sions (i.e., Models of Self and Others) were relatively
distinct from each other. This pattern of findings is con-
sistent with Bartholomew and Shaver’s (1998) findings
that adult attachment measures assess a core set of re-
lational tendencies, but that domain-specific attachment
patterns (e.g., romantic vs. close relationships) can be
substantially different. Given the attachment measures
were not redundant with each other, investigating the re-
lationships between both types of attachment and AS was
warranted.
Most participants (39%) in this study were classi-
fied as having a secure attachment style with the re-
mainder distributed relatively equally across the three
insecure attachment dimensions. These proportions are
consistent with earlier findings using the same measures
(e.g., Bartholomew & Horowitz, 1991; Brennan & Shaver,
1998), but contrast with those of Weems et al. (2002) who
reported fewer individuals with secure and dismissing at-
tachment (33 and 10%, respectively) and more individuals
with fearful and preoccupied attachment (25 and 32%,
respectively) in their college sample. These differences
might be attributable to demographic differences, such
as age or ethnicity. Weems et al.’s undergraduate sample
was slightly older than the present sample (21.7 years vs.
Tab l e V. Summary of Hierarchical Regression Analyses Predicting ASI-Physical Scores
Step Model R2Change in R2Fp<β
1. Age .00 .00 0.17 ns −.03
1. Gender .06 .06 13.56 .001 .26
2. STAI-T .15 .09 20.71 .001 .29
3. ECR Model of Self .22 .07 19.03 .001 .32
4. ECR Model of Self ×STAI-T .23 .01 1.29 ns −.41
3. ECR Model of Others .14 .00 0.28 ns −.04
4. ECR Model of Others ×STAI-T .15 .00 0.39 ns .00
3. RQ-RSQ Model of Self .18 .03 6.88 .01 −.20
4. RQ-RSQ Model of Self ×STAI-T .18 .00 0.72 ns .22
3. RQ-RSQ Model of Others .16 .01 3.29 ns .12
4. RQ-RSQ Model of Others ×STAI-T .16 .00 0.26 ns −.12
Note. ASI =Anxiety Sensitivity Index; STAI-T =State Trait Anxiety Inventory—Trait Scale; ECR =
Experiences in Close Relationships Inventory; RQ-RSQ =Combined zscores for the Relationship.
Questionnaire and Relationship Scales Questionnaire.

P1: IZO
joba2004.cls (03/17/2004 v1.1 LaTeX2e JOBA document class) pp1397-joba-NY00000635 October 9, 2004 20:22
198 Watt, McWilliams, and Campbell
Tab l e V I. Summary of Hierarchical Regression Analyses Predicting ASI-Psychological Scores
Step Model R2Change in R2Fp<B
1. Age .00 .00 1.07 ns .07
1. Gender .01 .01 0.14 ns .03
2. STAI-T .32 .31 99.33 .001 .56
3. ECR Model of Self .35 .03 7.88 .01 .19
4. ECR Model of Self ×STAI-T .35 .00 2.52 ns .52
3. ECR Model of Others .35 .03 10.59 .001 .19
4. ECR Model of Others ×STAI-T .36 .01 3.17 ns .48
3. RQ-RSQ Model of Self .33 .01 2.40 ns −.11
4. RQ-RSQ Model of Self ×STAI-T .34 .02 1.83 ns −.32
3. RQ-RSQ Model of Others .32 .00 0.01 ns −.01
4. RQ-RSQ Model of Others ×STAI-T .33 .01 0.03 ns .04
Note. ASI =Anxiety Sensitivity Index; STAI-T =State Trait Anxiety Inventory—Trait Scale; ECR =
Experiences in Close Relationships Inventory; RQ-RSQ =Combined zscores for the Relationship
Questionnaire and Relationship Scales Questionnaire.
19.4 years), and older individuals in the present sample
were significantly more likely to be classified as having
a fearful attachment style. In addition, most of Weems
et al.’s participants were Hispanic (61%) and African
American (19%), whereas participants in the present sam-
ple were predominantly Euro Canadian (98%).
Consistent with both attachment theory and Weems
et al.’s (2002) research, individuals classified as being ei-
ther securely attached or dismissing reported lower levels
of AS than individuals classified as either fearful or pre-
occupied. The pattern of results was similar across both
attachment measures but, with the measure of close rela-
tionships, differences in ASI scores between the fearful
group and both the secure and dismissing groups were
not statistically significant. Consistent with the roman-
tic attachment findings, the close relationship Model of
Self was strongly associated with AS and its lower-order
components, whereas the associations involving Model of
Others were more limited.
It had been hypothesized that attachment beliefs
regarding close relationships in general might be more
salient than romantic attachment beliefs when consider-
ing expectancies regarding the amount of assistance others
might provide during an episode of anxiety. The similar-
ity of the findings across attachment measures did not
support this hypothesis. However, one noteworthy dif-
ference was observed. The correlation analyses revealed
negative views of others in romantic relationships were
associated with fears of the psychological symptoms of
anxiety (i.e., AS-Psychological Concerns). In contrast,
when close relationship attachment was considered, a neg-
ative view of others was associated with fears of physical
anxiety symptoms (i.e., AS-Physical Concerns). These
findings suggest the possibility that within the intimacy
Table VII. Summary of Hierarchical Regression Analyses Predicting ASI-Social Scores
Step Model R2Change in R2Fp<B
1. Age .00 .00 0.07 ns .02
1. Gender .00 .00 0.21 ns .03
2. STAI-T .03 .03 7.24 .01 .18
3. ECR Model of Self .07 .04 8.54 .01 .23
4. ECR Model of Self ×STAI-T .07 .07 0.26 ns −.20
3. ECR Model of Others .04 .01 0.62 ns .06
4. ECR Model of Others ×STAI-T .04 .01 0.17 ns .13
3. RQ-RSQ Model of Self .06 .03 4.94 .05 −.18
4. RQ-RSQ Model of Self ×STAI-T .06 .00 1.20 ns .30
3. RQ-RSQ Model of Others .03 .00 0.26 ns .04
4. RQ-RSQ Model of Others ×STAI-T .04 .01 0.09 ns −.08
Note. ASI =Anxiety Sensitivity Index; STAI-T =State Trait Anxiety Inventory—Trait Scale; ECR =
Experiences in Close Relationships Inventory; RQ-RSQ =Combined zscores for the Relationship
Questionnaire and Relationship Scales Questionnaire.

P1: IZO
joba2004.cls (03/17/2004 v1.1 LaTeX2e JOBA document class) pp1397-joba-NY00000635 October 9, 2004 20:22
Anxiety Sensitivity and Attachment 199
of a romantic relationship the more private and less visi-
ble components of anxiety may be more salient; whereas,
within close relationships in general individuals may dis-
trust others to respond appropriately to the more visi-
ble manifestations of anxiety. It is noteworthy that one’s
perception of others may evoke different concerns (e.g.,
AS-Psychological vs. AS-Physical) in different contexts
(romantic vs. general relationships). This apparent speci-
ficity of AS components to particular contexts of learn-
ing has been demonstrated in other studies. For example,
both MacPherson, Stewart, and McWilliams (2001) and
Watt and Stewart (2003) found AS-Psychological Con-
cerns to play a unique role in relations between parental
drinking behavior in childhood and anxiety outcomes in
adulthood.
Weems et al. (2002) found the Model of Self dimen-
sion to predict AS levels after controlling for both current
anxiety symptoms and general psychological distress as
measured by the SCL-90. To examine this further, this
study investigated the role of trait anxiety as measured by
the STAI-T. Results were consistent with Weems et al.’s
(2002). For both attachment measures, a unique relation-
ship between the attachment dimensions and AS levels
was identified beyond what could be attributed to a gen-
eral propensity to experience fear or negative affect. For
both the romantic attachment and close relationships mea-
sures, the Model of Self dimension remained significantly
associated with ASI total scores and all three subscales
after accounting for the variance contributed by trait anx-
iety. However, for both measures, only one of the partial
correlations involving the Model of Others dimension re-
mained significant with approximately 6% of the variance
in AS-Psychological Concerns accounted for by the ECR
Model of Others dimension and only 2% of the variance
in AS-Physical Concerns scores accounted for by the cor-
responding RQ-RSQ dimension.
Finding that attachment variables account for unique
variance in AS beyond that contributed by measures of
anxiety has now been demonstrated in separate studies
utilizing two different operationalizations of anxiety. The
consistency of findings across the two studies was not
surprising given the similarity between state (i.e., current
symptoms) and trait anxiety. The results of these studies
provide support for the important role of interpersonal
factors (e.g., attachment characteristics) in shaping ex-
pectancies about one’s ability to manage stress, including
stress related to somatic sensations. Assuming continu-
ity between childhood and adult attachment styles, these
findings raise the possibility that negative experiences
that play a role in the development of insecure attach-
ment also contribute to the development of elevated AS
levels.
The relationship between Model of Others and AS
levels was of particular interest in this study. Weems et al.
(2002) had initially hypothesized that a negative Model
of Others could play a role in the development of high
AS but their findings provided little support for this idea.
For example, the Model of Others variable was not sig-
nificantly associated with ASI scores in their high school
sample and accounted for only 2.5% of the variance in
ASI scores in their university sample. However, they did
find the interaction between Model of Others and anxiety
to significantly predict ASI scores, raising the possibility
that this dimension could have a particular role to play
in the development of AS among more highly anxious
individuals. This study attempted to replicate this finding
and to investigate whether more general (vs. romantic)
attachment perceptions would be more salient when con-
sidering the amount of assistance others might provide
during an episode of anxiety. Results, however, revealed
no significant interaction between Model of Others and
trait anxiety in predicting ASI total or subscale scores
with either attachment measure.
Several limitations of this study need to be con-
sidered when interpreting the results. First, the cross-
sectional nature of the study precludes drawing conclu-
sions about the direction of relations between attachment
experiences and the development of AS. Prospective lon-
gitudinal research is required to determine whether inse-
cure attachment, in particular attachment characterized by
a negative Model of Self, is predictive of high AS or in-
stead that elevated levels of AS foster insecure attachment
beliefs. Although the focus of this study has been on the
former relationships, it is also possible that children with
heightened sensitivities to bodily sensations might com-
promise the attachment process by demanding more at-
tention and parental resources than less sensitive children.
Second, the attachment dimensions accounted for 24% of
the variance in ASI total scores, but the multiple regres-
sion analyses indicated that the attachment dimensions
accounted for a much smaller portion of unique variance
in ASI scores (i.e., 3–9% depending on the attachment
measure and ASI score considered). These modest effect
sizes suggest the need for caution before overstating the
role of attachment beliefs in the development of high AS.
Nonetheless, finding a role for attachment is consistent
with the growing body of literature attesting to the impor-
tance of early childhood experiences in the development
of AS (e.g., Watt et al., 1998; Watt & Stewart, 2000, 2003).
Further research is required to determine the extent of that
role and to further delineate the relative roles of one’s per-
ception of self versus one’s perception of others. To date,
studies have investigated relations between AS and attach-
ment in nonclinical samples only. Replicating the findings

P1: IZO
joba2004.cls (03/17/2004 v1.1 LaTeX2e JOBA document class) pp1397-joba-NY00000635 October 9, 2004 20:22
200 Watt, McWilliams, and Campbell
from these studies with clinical samples would enhance
external validity and could inform treatment approaches to
anxiety-related and other disorders for which high AS is a
risk.
In summary, this study successfully replicated and
extended the findings of Weems et al. (2002). Attachment
insecurity in both romantic and nonromantic relation-
ships, particularly insecurity characterized by a negative
Model of Self (fearful and preoccupied), was associated
with elevated levels of AS and AS components. The Model
of Others played a more limited role in relation to AS. This
study provides further support for the role of interpersonal
factors in the development of AS and suggests the need for
continued investigation into parental practices and child-
hood learning experiences in the development of such risk
factors for psychopathology.
REFERENCES
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing andA1
interpreting interactions. Newbury Park, CA: Sage.
Bartholomew, K., & Horowitz, L. M. (1991). Attachment styles among
young adults: A test of a four-category model. Journal of Person-
ality and Social Psychology,61(2), 226–244.
Bartholomew, K., & Shaver, P. R. (1998). Methods of assessing adult
attachment: Do they converge? In J. A. Simpson & W. S. Rholes
(Eds.), Attachment theory and close relationships. New York: Guil-
ford.
Bowlby, J. (1988). The origins of attachment theory. In J. Bowlby (Ed.),
A secure base: Parent child attachment and healthy human devel-
opment (pp. 20–38). New York: Basic Books.
Brennan, K. A., Clark, C. L., & Shaver,P. R. (1998). Self-report measure-
ment of adult attachment: An integrative interview. In J. A. Simpson
&W.S.Rholes(Eds.),Attachment theory and close relationships
(pp. 46–76). New York: Guilford Press.
Brennan, K. A., & Shaver, P. R. (1998). Attachment styles and per-
sonality disorders: Their connections to each other and to parental
divorce, parental death, and perceptions of parental caregiving.
Journal of Personality,66(5), 835–878.
Cox, B. J., Borger, S. C., & Enns, M. W. (1999). Anxiety sensitivity and
emotional disorders: Psychometric studies and their theoretical im-
plications. In S. Taylor (Ed.), Anxiety sensitivity: Theory, research,
and treatment of the fear of anxiety (pp. 115–148). Mahwah, NJ:
Erlbaum.
Derogatis, L. R. (1983). Symptom Checklist-90: Administration, scoring
and procedures manual-II. Towson, MD: Clinical Psychometric
Research.
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(3), 430–
445.
Hazan, C., & Shaver, P. (1987). Romantic love as conceptualized as an
attachment process. Journal of Personality and Social Psychology,
52, 511–524.
MacPherson, P. S. R., Stewart, S. H., & McWilliams, L. A. (2001).
Parental problem drinking and anxiety disorder symptoms in adult
offspring: Examining the mediating role of anxiety sensitivity com-
ponents. Addictive Behaviors,26, 917–934.
Mickelson, K. D., Kessler, R. C., & Shaver, P. R. (1997). Adult attach-
ment in a nationally representative sample. Journal of Personality
and Social Psychology,73, 1092–1106.
Peterson, R. A., & Reiss, S. (1992). Anxiety Sensitivity Index Man-
ual (2nd ed.). Worthington, OH: International Diagnostic Systems
Reiss (1991).
Reiss, S. (1991). Expectancy theory of fear, anxiety, and panic. Clinical
Psychology Review,11, 141–153.
Rutter, M. (1997). Clinical implications of attachment concepts: Ret-
rospect and prospect. In L. Atkinson & K. J. Zucker (Eds.), At-
tachment and psychopathology (pp. 17–36). New York: Guilford
press.
Schmidt, N. B., Lerew,D. R., & Jackson, R. J. (1997). The role of anxiety
sensitivity in the pathogenesis of panic: Prospective evaluation of
spontaneous panic attacks during acute stress. Journal of Abnormal
Psychology,106, 355–364.
Silverman, W. K., & Weems,C. F. (1999). Anxiety sensitivity in children.
In S. Taylor (Ed.), Anxiety sensitivity: Theory, research, and treat-
ment of the fear of anxiety (pp. 239–268). Mahwah, NJ: Erlbaum.
Spielberger, C. D., Gorsuch, R. L., Lushene, R. E., Vagg, P. R., & Jacobs,
G. A. (1983). Manual for the State-Trait Anxiety Inventory. Palo
Alto, CA: Consulting Psychologists Press.
Stein, M. B., Jang, K. L., & Livesley, W. J. (1999). Heritability of
anxiety sensitivity: A twin study. American Journal of Psychiatry,
156, 246–251.
Taylor, S. (1995). Anxiety sensitivity: Theoretical perspectives and
recent findings. Behaviour Research and Therapy,33, 243–
258.
Watt, M., Stewart, S. H., & Cox, B. J. (1998). A retrospective study of
the origins of anxiety sensitivity. Behaviour Research and Therapy,
36, 505–525.
Watt, M. C., & Stewart, S. H. (2000). Anxiety sensitivity mediates the
relationships between childhood learning experiences and elevated
hypochondriacal concerns in young adulthood. Journal of Psycho-
somatic Research,49, 107–118.
Watt, M. C., & Stewart, S. H. (2003). The role of anxiety sensitivity com-
ponents in mediating the relationship between childhood exposure
to parental dyscontrol and anxiety symptoms in adulthood. Journal
of Psychopathology and Behavioral Assessment,25, 167–176.
Weems, C. F., Berman, S. L., Silverman, W. K., & Rodriquez, E. T.
(2002). The relation between anxiety sensitivity and attachment
style in adolescence and early adulthood. Journal of Psychopathol-
ogy and Behavioral Assessment,24, 159–168.
Zinbarg, R. E., Mohlman, J., & Hong, N. N. (1999). Dimensions of
anxiety sensitivity. In S. Taylor (Ed.), Anxiety sensitivity: The-
ory, research, and treatment of the fear of anxiety (pp. 83–113).
Mahwah, NJ: Erlbaum.

P1: IZO
joba2004.cls (03/17/2004 v1.1 LaTeX2e JOBA document class) pp1397-joba-NY00000635 October 9, 2004 20:22
Queries to Author:
A1: Pub: Please note that this reference is not cited in the text.
201

P1: IZO
joba2004.cls (03/17/2004 v1.1 LaTeX2e JOBA document class) pp1397-joba-NY00000635 October 9, 2004 20:22
Queries to Publisher:
P1: Pub: Please provide article accepted date.
202