Content uploaded by Mario Mikulincer
All content in this area was uploaded by Mario Mikulincer on Dec 01, 2015
Content may be subject to copyright.
Attachment security and pain — The disrupting effect of captivity
Tonny Elmose Andersen
, Ruth Defrin
, Mario Mikulincer
, Zahava Solomon
Department of Psychology, University of Southern Denmark, Denmark
Bob Shapell School of Social Work, Tel Aviv University, Israel, Postdoc, University of Southern Denmark, Denmark
School of Health Professions, Tel Aviv University, Israel
New School of Psychology, Interdisciplinary Center (IDC), Herzliya, Israel
Received 14 January 2015
Received in revised form 1 September 2015
Accepted 24 October 2015
Available online xxxx
The present study assesses the possible disruption effect of posttraumatic stress symptoms (PTSS) with regard to
the protective role of attachment on pain, among ex-POWs. While secure attachment seems to serve as a buffer,
decreasing the perception of pain, this function may be disrupted by PTSS. The study sample included 104 sub-
jects who were combat veterans of the 1973 Yom Kippur War comprising of 60 male ex-prisoners of war (ex-
POWs) and 44 comparable male combat veterans. Both attachment and pain were investigated experimentally
in the laboratory and via questionnaires. We found that ex-POWs showed higher levels of clinical pain and at-
tachment insecurities compared to controls. Moreover, attachment avoidance and soothing effect of attachment
(SEA) were both associated with lower levels of clinical pain. Most importantly, PTSS moderated the associations
between attachment and pain, as well as the mediation role of attachment between captivity and pain. The re-
sults imply that although attachment can be an important resource for coping with pain, it can be severely
disrupted by PTSS among trauma survivors.
© 2015 Published by Elsevier Inc.
War captivity is an extreme traumatic experience that entails harsh
physical and psychological abuse, such as torture, deprivation of basic
needs, humiliation, threats and mock executions . Even decades
after repatriation, captivity often has long-term psychiatric (i.e., PTSD),
somatic and interpersonal ramiﬁcations [e.g.,2,3]. The present study in-
vestigates the relationship between attachment, pain, and posttraumat-
ic stress symptoms (PTSS) among ex-POWs and control veterans.
Pain, attachment and war captivity
Physical abuse during captivity might negatively affect pain [4,5].
Ex-POWs have been found to experience high levels of persistent clini-
cal pain and altered pain perception [4,5]. Furthermore, research indi-
cated that ex-POWs had dysf unctional pain modulation and higher
levels of pain reporting when their pain threshold was reached [4,5].
The emotional abuse asso ciated with captivity might impact the
ability to trust others, manifested in attachment [e.g.,1]. Attachment ori-
entations are cognitive-emotional schemas of the self and others .
Adult attachment can be conceptualized along two dimensions: attach-
ment anxiety (worry over the availability and positive regard of others)
and attachment avoidance (discomfort with closeness and dependence
on others) . Low levels on both dimensions indicate attachment
Attachment may change as a result of meaningful interpersonal life
events . While positive life events are associated with attachment se-
curity , negative life events are implicated with attachment insecurity
. As such, exposure to prolonged and deliberate interpersonal trau-
ma, such as captivity, may undermine attachment . Indeed, studies
[11,12] revealed higher attachment insecurities, which increased over
time among ex-POWs, compared to decreasing or unchanging attach-
ment insecurities among controls.
Pain, attachment and PTSS
The mutual maintenance model  suggests that PTSD and pain are
mutually maintained by a variety of psychological mechanisms, includ-
ing elevated arousal, attention bias, reasoning bias, reminders of the
trauma, and avoidant behaviours. Indeed, research indicates that PTSS
are associated with higher clinical pain in both civilian and military
samples [15–17]. Furthermore, traumatiz ed indivi duals experienced
more intense and widespread clinical pain, compared to healthy con-
trols and patients with anxiety . A more compl ex pattern was
found with regard to experimental pain [14-16]. For example, a previ-
ous study indicated that traumatized individuals had higher pain
thresholds (i.e., decreased sensitivity to pain). However, at the same
Journal of Psychosomatic Research 79 (2015) 471–476
⁎ Corresponding author at: Department of Psychology, University of Southern Denmark,
Campusvej 55, DK-5230 Odense M, Denmark.
E-mail address: firstname.lastname@example.org (T.E. Andersen).
0022-3999/© 2015 Published by Elsevier Inc.
Contents lists available at ScienceDirect
Journal of Psychosomatic Research
time they reported elevated pain when the thresholds were exceeded
Literature regarding PTSS and attachment implies a reciprocal rela-
tionship. The ﬁrst direction implies that attachment might predict
PTSS. Ex posure to trauma activates attachment by recalling ﬁgures
that provide a sense of protection . This mobilization might act as
a resilience factor, helping the indivi dual co pe and decrease the risk
for PTSS. Moreover, even after the trauma, activation of the attachment
system might ease the negative effects of PTSS, creating a soothing effect
. In line with this contention, a prospective study among Israeli un-
dergraduates indicated a relationship between attachment before the
Gulf War and the subsequent development of PTSS. Furthermore, re-
search indicated a soothing effect of attachment (SEA), i.e., positive ef-
fect of secure representations on cognitive manifestation of PTSD .
Participants with PTSD had longer response times in id entifying the
color of trauma-r elated words, compared to negative or neutral
words, indicating greater mental accessibility of the traumatic content
. However, priming with a security-related word reduced this
heightened accessibility .
However, in the second direction of association, PTSS might detri-
mentally affect attachment. Symptoms such as avoidance or outbursts
of anger, may lead to negative interpersonal events, which may height-
en attachment insecurities . Furthermore, it is suggested that while
a healthy attachment system seems to serve as a buffer against emo-
tional and physical stressors, it may be disrupted by traumatic reaction
.Speciﬁcally, Herman observed that prolonged captivity harms
human relationships and leads to dysregulation of interpersonal func-
tioning . Likewise, Mikuli ncer et al.  noted that the reactivation
of trauma, re ﬂected in PTSS, might erode the sense of security and
lead to disorganization of the attachment system over time, so that its
functioning might be damaged.
Research supports both possibilities. Studies among ex-POWs found
that PTSS predicted increased attachment insecurities over time [11,12].
Furthermore, ex-POWs with chronic PTSD exhibited disruptions in the
soothing and healing functions of attachment . Although ex-POWs
searched for representations of a security provider, this activation of at-
tachment failed to soothe them .
Attachment and pain
According to Bowlby , attachment motivates seeking proximity to
signiﬁcant others as a way of attaining security and comfort when in
pain. The relationship between attachment and pain might also rely
on affect regulation. It has been proposed that attachment insecurities
might be related to persistent pain due to dysfunctional affect regula-
tion and problematic adjustment to acute pain . Research indicates
associations between attachment insecurities and persistent clinical
pain [20-23]. A number of experimental pain studies found attachment
insecurities to be associated with both lower [23-26] and higher pain
Although research on the effect of secure representations on pain is
quite limited, studies suggest that attachment may have a buffering ef-
fect on the experience of pain. Pictures of attachment ﬁgures, compared
to pictures of strangers, have found to lead to reductions in pain and
pain related neural activity . Likewise, priming mental representa-
tions of loved ones had a positive effect, leading to higher pain
SS, captivity, attachment and pain
The documented associations between attachment and pain [e.g.,23,
27], as well as PTSS and pain [e.g.,17] raise two questions. The ﬁrst
regards the mediation role of attac hment within the association be-
tween captivity and pain. Speciﬁcally, whether attachment acts as a
mechanism, thereby explaining increased pain among ex-POWs. Expo-
sure to captivity might damage attachment, leading to high attachment
insecurities and low SEA. These negative effects on attachment may, in
turn, predict ampliﬁed pain.
The second question regards the moderation role of PTSS between
captivity, attachment and pain. While attachment as a buff er against
pain is well documented, it is still unclear whether this function of at-
tachment may be disrupted by traumatic response, i.e., PTSS. Speciﬁcal-
ly, PTSS may not only lead to increased attachment insecurities, it may
also harm the protective function of attachment on pain. Hence, while
attachment security and SEA might be associated with lower pain and
act as a mechanism to help relieve the negative effects of captivity on
pain, this process might be cancelled or overshadowed by the negative
impact of high PTSS.
To the best of our knowledge both the mediation role of attachment
between captivity and pain, as well as the moderation role of PTSS be-
tween captivity, attachment and pain, have not been investigated. Ex-
amination of these signiﬁcant issues has both important theoretical
and clinical contributions. The present study aimed to ﬁll these knowl-
edge gaps by exploring these issues among ex-POWs and controls.
We hypothesized: A) ex-POWs, compared to controls, will experi-
ence higher pain threshold, levels of clinical pain and attachment inse-
curities and lo wer levels of SEA; B) PTSS will be associated with
higher pain threshold, levels of clinical pain and attachment insecurities
and lower levels of SEA; C) high attachment insecurities and low SEA
will be associated with higher levels of clinical pain and lower pain
threshold; D) PTSS will moderate between attachment insecurities
and SEA, and the clinical level of pain and pain threshold; and E) PTSS
will moderate the mediating role of attachment insecurities and SEA be-
tween captivity, clinical pain and pain threshold.
The present study uses data from a longitudinal study on the psycho-
logical implications of war . The sample included 104 Israeli combat
veterans of the 1973 Yom Kippur War: 60 ex-POWs (mean age 57.5 ±
3 years) and 44 control combat veterans (58.8 ± 4 years).
According to Israel's Ministry of Defense, 240 combat soldiers were
POWs during the Yom Kippur War. The ex-POW groups consisted of
veterans who were captured and imprisoned in Egypt or Syria. The du-
ration of captivity in Egypt was 6 weeks (44/60 of the present sample)
and 9 months in Syria (16/60) from 1973 to 1974. In previous measures
the ex-POW group included 164 participants in 1991, 144 in 2003, and
183 in 2008.
The control group included combat veterans from the same war,
who were never held captive. Co ntrols were selected on the basis of
similarity to the ex-POWs in regard to military and personal variables
such as age, combat exposure, and rank. In previous measures the con-
trol group included 185 participants in 1991, 143 in 2003, and 118 in
Overall 1–17.5% data were missing across the study variables. No dif-
ferences were found b etween the 2 groups with regard to religion
(2) = 1.82, n.s.), income (t(92) = 1.58, n.s.), age (t(100) = −.28,
n.s.) and health before the war (χ
(1) = .28, n.s.). Twenty-six percent
of participants were married during the war and 70% had completed
Participants were obtained through the records of the Israeli Minis-
try of Defense. We sent letters of invitation to participate in the present
study, followed by a phone call two weeks later. Those who agreed to
participate in the study were scheduled for a single testing session.
Each participant signed an informed consent after receiving a detailed
explanation on th e study's aims and procedures. The stu dy was ap-
proved by the institutional review board of Tel-Aviv University.
472 T.E. Andersen et al. / Journal of Psychosomatic Research 79 (2015) 471–476
Materials and procedure
Heat–pain threshold was used to evaluate the sensitivity of the pain
system. The threshold was measured using a Peltier-based computer-
ized thermal stimulator (TSA II, Medoc Ltd., Ramat Yishai, Israel), with
a 3 × 3 cm contact probe. The probe of the stimulator was attached to
the forearm with a Velcro strap. Subjects rec eived four successive
ramps of gradually increasing temperature, starting from a baseline of
35 °C, at a rate of 2 °C/s with an inter-stimulus interval of 30 s. During
each ramp, subjects were asked to press a switch when a pain sensation
was ﬁrst perceived, thus deﬁning heat–pain threshold. Pressing the
switch resulted in an automatic recording of the threshold temperature
and reset the probe temperature to baseline. Heat–pain threshold was
computed by averaging the readings of four successive trials .
The McGill pain questionnaire
The McGill pain questionnaire (MPQ; ) provides a quantitative
evaluation of the patient's clinical pain experience with a separate mea-
sure of sensory, affective and cognitive dimensions. We used two quan-
titative parameters from the MPQ: pain rating index (PRI) — the total
sum of the values of words chosen out of a list of pain descriptors and
number of words chosen (NWC) — the number of words chosen from
SEA was investigated experimentally in the laboratory. First, partic-
ipants were asked to provide the ﬁrst names of secure ﬁgures using
the 6-item WHOTO scale , i.e., people from whom they sought prox-
imity (2 items), provided them with a safe haven (2 items), and provid-
ed them with a secure foundation for engaging in other activities (2
items). Second, they provided the ﬁrst names of their parents, siblings,
grandparents, close friends, and current and previous romantic part-
ners. Third, we read them a list of 100 ﬁrst names and asked them to in-
dicate the names of people they knew (acquaintances). This built a list
of names for each participant of security providers, other close persons
not includ ed in the WHOTO scale, and acquaintan ces. These names
were used in the computerized tasks. No participant had difﬁculty nam-
ing security providers.
We used Stroop color-naming procedure  to examine the effect
of secure representations in reducing the accessibility of trauma-
related mental representations. This procedure has been used in previ-
ous studies . The target words were either war-related (e.g., captivity,
prison), negatively valenced (e.g., povert y, failure), or neutral
(e.g. , circle, dress). Words were printed in H ebrew and matched for
the number of letters, syllables, and frequency of use. Different blocks
of ex perimental trials were preceded by subliminal presentations of
the name of a participant's attachment ﬁgure  or the name of an ac-
quaintance. Following 30 practice trials, participants performed 144 ex-
perimental trials. Each experimental trial began with an X in the middle
of the screen, which was replaced by a rapid (32-millisecond) sublimi-
nal presentation of one of 4 prime names (i.e., attachment ﬁgure, close
friend, acquaintance, unknown) followed immediately by a masking
stimulus (an XXX pa ttern) for 500 ms. Immediately following the
prime, one of six possible words (2 captivity-related words, 2 negatively
valenced words, 2 neutral words) was presented for 1000 ms in 1 of 4
colors. Each combination of name and color was presented three
times with each of the two primes (12 trials for each word in each prim-
ing condition = 24 total trials for each word), for a total of 144 trials.
None of the participants perceived the subliminal presentation of the
was measured as the response time for color-identifying war-
related words after attachment ﬁgure name-priming compared to the
response time for the same words after other-priming (i.e. , name of
close frien d, acquaintance or unknown). The lower the gap (i.e., the
lower the response time after attachment ﬁgure name-priming while
controlling for the response time after other-priming), the stronger
the soothing effect.
Attachment anxiety and avoidance
Attachment anxiety and avoidance were assessed using the 36-
items Experiences in Close Relationships questionnaire (ECR; ). Par-
ticipants were asked to think about their close relationships, without fo-
cusing on a speciﬁc pers on, and rate the exte nt to which each item
accurately described their feelings in such relationships using a
7-point scale ranging from 1 (not at all) to 7 (very much). Eighteen
items measured attachment-related anxiety (e.g., I worry about being
abandoned) and 1 8 items measured avoida nce (e.g., I prefer not to
show a partner how I feel deep down). The reliability and validity of
the two subscales have been repeatedly demonstrated [e.g.,6]. In the
present study, the inventory was found to have high internal consisten-
cy (Cronbach's α = 0.89, 0.91 for attachment avoidance and attachment
The PTSD Inventory
The PTSD Inventory  was used to measure combat-related PTSS.
This instrument is based on the clusters of PTSS according to DSM-IV
. The questionnaire consists of 17 statements describing different
expressions of the disorder following war experience. Respondents
rate each statement according to the frequency of the experience during
the last month, on a 4-point scale from ‘never’ to ‘very often’.Thetotal
score for the scale was based on the total number of symptoms en-
dorsed. In the present study, the inventory was found to have high in-
ternal consistency (Cronbach's α = 0.96). The scale was also found to
have high convergent validity when compared to diagnoses based on
structured clinical interviews .
Inter-correlations between the main study measures are presented
in Table 1.
Hypothesis A. Captivity, attachment and pain.
To examine differences in pain and PTSS as a function of type of trau-
ma (ex-POWs versus combat veterans), we conducted a series of t-tests
for in depen dent samples with type of trauma as factor, and pain
Inter-correlations between the main study measures.
Measure 1 2 3 4567
1 .09 .11 −.06 .16 .07 .00
2. NWC 1 .95
3. PRI 1 .41
4. PTSS 1 .36
7. SEA 1
M (SD) 45.10
Notes. 1. PTSS = posttraumatic stress symptoms; SEA = soothing effect of attachment.
SEA was measured in milliseconds. Pearson correlations are computed as case-wise rela-
tionships. Coefﬁcients for SEA and other variables represent partial correlations.
= p b .05.
= p b .01.
= p b .001.
473T.E. Andersen et al. / Journal of Psychosomatic Research 79 (2015) 471–476
measures (PRI, NWC and pain threshold) as well as PTSS as dependent
variables. To examine differences in attachment as a function of type
of trauma we conducted multivariate analysis of variance (MANOVA)
with type of trauma as factor and the two dimensions of attachment in-
security measures (attachment avoidance, attachment anxiety) as the
dependent variables. Lastly, we conducted analysis of covariance
(ANCO VA) with type of trauma as factor, response time for war-
related words after attachment ﬁgure name-priming as dependent var-
iable and response time for war-related words after other-priming as
covar iates, in order to examine SEA (see Table 2). We controlled for
Type I error with Bonferroni corrections.
Analysis of PRI and PTSS revealed signiﬁcant effect for type of trau-
ma. Ex-POWs reported higher pain ratings and more PTSS compared
to combat veterans. Analysis of NWC and pain threshold did not reveal
asigniﬁcant effect for type of trauma. Analysis on trauma-SEA did not
reveal signiﬁcant effect for type of trauma.
MANOVA on attac hment avoidance and attac hment anxiety re -
vealed signiﬁcant effect for type of trauma, Pillai's Trace F(4180) =
5.44, p b .001, Partial Eta
= .11. Ex-POWs reported higher levels of at-
tachment avoidance and attachment anxiety compar ed to combat
Hypothesis B. PTSS, attachment and pain.
In order to examine the relationship between PTSS on the one hand,
and pain measures and attachment insecurities, on the other, we con-
ducted two-tailed Pearson correlations. In order to examine the rela-
tionship between PTSS and SEA, we conducted partial correlations
with PTSS as the independent variable, response time for war-related
words after attachment ﬁgure name-priming as the dependent variable,
and response time for war-related words after other-priming as covar-
iates (see Table 1).
The analyses revealed signiﬁcant positive correlations between PTSS
and PRI as well as NWC. Meaning, the more the PTSS, the more pain de-
scriptor words chosen and the higher the pain rating. Analyses revealed
signiﬁcant positive correlations between PTSS and attachment insecu-
rities. The more the PTSS, the higher the attachment insecurities. The re-
lationship between PTSS and pain threshold as well as SEA was non-
Hypothesis C. Attachment and pain.
In order to examine the relationship between attachment insecu-
rities and pain measures we conducted series of linear regressions
with attachment insecurities as the independent variabl es and pain
measures as dependent variables. In order to examine the relationship
between SEA and pain measures we conducted a series of linear regres-
sions with response time for war-related words after attachment ﬁgure
name-priming as the independent variable, and pain measures as the
dependent variables. Response time for war-related words after other-
priming was used as covariate.
The relationships between attachment avoidance and attachment
anxiety and PRI (β = .20, p = .07; β = .09, p = .39, respectively), as
well as pain threshold (β =.15,p =.14;β =.02,p = .83, respectively)
were non-signiﬁcant. However, while the relationship between attach-
ment anxiety and NWC was non-signiﬁcant (β = .12, p =.26),there-
lationship between atta chment avoidance and NWC was signiﬁcant
(β = .21, p = .05); higher attachment avoidance implied more pain de-
scriptor words and a higher pain rating.
The relationship between SEA and pain thre shold was non-
signiﬁcant (β =.00,p = .99). However, th e rela tionships between
SEA and PRI as well as NWC were signiﬁcant (β =.66,p =.02;β =
.70, p = .02, respectively). The higher the response time for war-
related words after attachment ﬁgu
re n ame-priming (i.e., the lower
the SEA), the more pain descriptor words chosen and a higher pain
Hypothesis D. PTSS as moderator between attachment and pain.
To examine Hypothesis D, we performed a series of multiple regres-
sion analyses for attachment insecurities, SEA and the interactions with
PTSS, predicting pain measures, and controlling for 2 study groups (ex-
POWs vs. controls). All of our predictors were mean-centered prior to
entering them into the regression a nalyses. Signiﬁcant interactions
were probed using the Hayes's  PROCESS (model 1) computational
macro. Table 3 presents the regression coefﬁcients.
The interaction between SEA and PTSS contributed signiﬁcantly to
the explained variance of pa in threshold. Simple e ffect analyse s re-
vealed that among participants with low PTSS there was a signiﬁcant ef-
fect of attachment soothing on the pain threshold (β = − 1.03, p =.02)
— the longer the response time for war-related words after attachment
ﬁgure name-priming (i.e., lower SEA), the lower the pain threshold.
However, among participants with high PTSS, the soothing effect on
pain threshold was not signiﬁcant (β =.29,p = .38). Other interactions
between SEA and PTSS in predicting pain measures were non-
The interaction between attachment avoidance and PTSS contribut-
ed signiﬁcantly to the explained variance of NWC. Simple effect analyses
Means, SDs, t tests/analyses of variance F results of pain and attachment measures among
ex-POWs and controls.
Ex-POWs Controls t(df)/F (df1, df2) Cohen's
n = 60 n = 44
M (SD) M (SD)
PRI 20.22 (10.66) 15.19 (11.49) − 2.08
NWC 8.98 (4.07) 7.22 (4.91) − 1.80 (83) .39
Pain threshold 45.30 (3.19) 44.83 (4.11) −.65
4.47 (1.08) 3.60 (.87) 19.42
3.37 (1.08) 2.56 (1.03) 14.56
SEA 942.43(292.90) 1000.50(264.47) 2.50 (1, 97) .03
PTSS 9.74 (4.37) 2.61 (4.27) − 7.90
Note. Ex-POWs = ex-prisoners of war; PRI = pain rating index; NWC = the number of
words chosen to describe pain; PTSS = posttraumatic stress symptoms; SEA = soothing
effect of attachment. SEA was measured in milliseconds.
= p b .05.
= p b .01.
= p b .001.
Hierarchical multiple regression analyses predicting pain measures with group, attach-
ment, PTSS, and the interaction between PTSS and attachment.
Pain threshold NWC PRI
Model 1 Group .09 −.12 −.04
Attachment anxiety .17 .05 −.00
Attachment avoidance .00 .13 .12
PTSS −.19 .45
Attachment anxiety × PTSS .05 .00 .05
Attachment avoidance × PTSS .06 −.27
Model 3 Group − 11 −.04 .03
PTSS .45 1.44
SEA −.40 .71
SEA × PTSS 2.654
Note: Ns ranged from 79 to 103. The table depicts a series of regression analyses. PRI =
pain rating index; NWC = the number of words chosen to descri be pain; PTSS =
posttraumatic stress symptoms. SEA = soothing effect of attachment. SEA was measured
= p b .05
= p b .01
= p b .001.
474 T.E. Andersen et al. / Journal of Psychosomatic Research 79 (2015) 471–476
revealed that among participants with low PTSS there was a signiﬁcant
effect of attachment avoidance on NWC (β = 2.30, p = .02) — the lower
the attachment avoidance, the fewer pain descriptor words. However,
among participants with high PTSS, the effect of attachment avoidance
on NWC was non-signiﬁcant (β =1.07,p =.22).
The interaction between attachment avoidance and PTSS contribut-
ed signiﬁcantly to the explained variance of PRI. Simple effect analyses
revealed that among participants with low PTSS there was a signiﬁcant
effect of attachment avoidance on PRI (β =.43,p = .04) — the lower the
attachment avoidance, the lower the pain rating. However, among par-
ticipants with high PTSS, the effect of attachment avoidance on PRI was
non-signiﬁ cant (β = −.21, p = .2 1). Other interactions were non-
Hypothesis E. The mediating role of attachment in the effect of captiv-
ity on pain — PTSS as a moderator.
We conducte d a moderated mediation model using PROCESS
(model 14) to investigate whether the level of PTSS moderated the me-
diation role of attachment within the relationship between type of trau-
ma and pain measures. The correlation between type of trauma and SEA
was non-signiﬁcant (r = .16, n.s.). Hence, we did not include SEA in the
analyses as a mediator. In addition, the correlation s between type of
trauma and pain threshold (r = .06, n.s.) as well as NWC were non-
signiﬁcant (r = .19, n.s.). Hence, we did not include pain threshold
and NWC in the analyses as outcome measures.
The a nalysis included type of trauma as the predictor, attachment
anxiety and attachment avoidance as the mediators, PRI as the out-
come measure, and PTSS as moderator. The model revealed t hat
only attachment avoidance mediated the relationship between
type of trauma and PRI. Type of trauma had a signiﬁcant main effect
on attachment avoidance (the “a” path in mediation analysis), b =
.26, p b .05. In addition, PTS S signiﬁcantly moderated the link be-
tween at tachment avoidance and PRI, b = −.31, p b .05. In other
words, the path (the
in mediation analysis) going from at-
tachment avoidance (i.e., the mediator) to PRI (i.e., t he outcome)
was qualiﬁed by PTSS. As expected, a higher degr ee of attachment
avoidance increased PRI among participants with low PTSS, but did
not predict PRI among participants with high PTSS. In other words,
being an ex-POW was associated with hi gher attachment avoidance,
which in turn increased PRI among participants with low PTSS (Boot-
strap conﬁdence i nterval (CI) [.02, .30]). This indirect effect was non-
signiﬁcant among participants with high PTSS (Bootstrap conﬁdence
interval (CI) [− .18, .02]).
The present study aimed to investigate the role of PTSS in the asso-
ciations between war captivity, attachment and pain. As expected, ex-
POWs reported higher levels of pain and attachment insecurities com-
pared to controls. The present ﬁndings are consistent with previous
studies [4,12], indicating the widespread and long-term ramiﬁcations
of war captivity. It is suggested that physical torture and psychological
abuse during captivity alter pain modulation and contribute to elevated
pain . At the same time, the abusive relationship with the captor
may permeate the captive's personality, thereby harming basic personal
resources, such as attachment .
Consistent with previous studies [e.g.,14] we found that PTSS were
associated with higher levels of clinical pain. According to the mutual
maintenance model, PTSS and pain are mutually maintained by psycho-
logical mechanisms, such as reminders of trauma and avoidant behav-
iours . However, another possibility is that physical harm resulting
from torture led to both PTSS and elevated pain. The present study did
not control for physical harm and hence cannot rule out this alternative.
Future studies should account for this while assessing the link between
PTSS and pain.
Results regarding the associations between PTSS and attachment in-
securities correspond with previous research [e.g.,18]. It might be that
elevated insecure attachment increases the risk for PTSS. However, it
is also possible that the negative interpersonal implications of PTSS pre-
dicted negative changes in attachment . Due to the nature of the
present study, the precise nature of the association between attachment
and PTSS cannot be determined. However, it is suggested that the rela-
tionship between the two variables is mutual.
As expected, signiﬁcant correlations between attachment avoidance
and SEA, on the one hand, and clinical pain, on the other hand, were
found. These results are consistent with previous studies linking attach-
ment insecurities to pain [e.g.,20]. According to the attachment theory,
attac hment acts as a resource . Hence, low levels of attachment
avoidance and high levels of SEA might impact coping with pain, leading
to decreased pain perceptions . An alternative explanation, which
arises from social pain theory , suggests that physical pain and social
distress are based on a shared neurological system. Hence, high attach-
ment avoidance or low SEA might lead to elevated physical pain .
Contrary to our hypotheses, we did not ﬁnd associations between at-
tachment anxiety and pain. This discrepancy between our ﬁndings
and existing theory and empirical ﬁndings [20,21] is unexpected.
The main contribution of the present research relates to the moder-
ing role of PTSS between attachmen t and pain. While attachment
avoidance and SEA were related with pain among participants with
low PTSS, these associations became non-signiﬁcant among participants
with high PTSS. Moreover, the relationship between captivity and pain
ratings was mediated by attachment avoidance only among individuals
with low PTSS. Our results imply that although attachment is an impor-
tant personal resource, its positive impact on pain may not be strong
enough among trauma survivors who suffer from elevated PTSS. One
explanation for the present ﬁndings offers that attachment does have
positive impact on pain among individuals with high PTSS. However,
this effect is obscured by the pervasive negative implications of higher
levels of PTSS on pain. An alternative explanation suggests that PTSS dis-
rupt the attachment system to such an extent ,thatitbecomesinef-
fective in reducing pain.
One may suggest that re-experiencing of the traumatic event, as
manifested by high PTSS, exposes the traumatized individual to the lim-
itations of close relationships as a source of safety. At these moments,
the protective ro le of attachment ﬁgures (real or internalized) is
undermined as the mental apparatus is ﬂooded. The traumatized victim
feels helpless and faces the failure of the attachment system to provide a
sense of security. Over time, this might erode the attachment system's
protective role against distress such as pain.
Several limitations may have affected our ﬁndings. First, the present
study uses cross-sectional data, hence we cannot infer the direction of
causality between attachment, pain and PTSS. Second, the present
study uses a highly selective sample of ex-POWs and combat veterans.
This emphasizes the necessity of caution in generalizing our results.
The present ﬁndings have important theoretical and clinical implica-
tions for the treatment of trauma survivors. Our results call attention to
the possible role of PTSS in the association between attachment and
pain. Ex-POWs who suffer from high levels of PTSS may be particularly
vulnerable to pain, partially due to damaged attachment. The present
results imply that although attachment can be an important resource
for coping with pain, it can be severely disrupted by PTSS. It is important
that clinical interventions will focus on the relief of PTSS before enhanc-
ing secure attachment.
The authors have no competing interests to report.
 J.L. Herman, Trauma and Recovery: The Aftermath of Violence, Basic Books, New
 M. Mikulincer, Z. Solomon, P. Shaver, T. Ein-Dor, Attachment-related consequences
of war captivity and trajectories of posttraumatic stress disorder: a 17-year longitu-
dinal study, J. Soc. Clin. Psychol. 33 (2014) 207–228.
475T.E. Andersen et al. / Journal of Psychosomatic Research 79 (2015) 471–476
 Y. Lahav, R. Rodin, Z. Solomon, Somatic complaints and attachment in former pris-
oners of war: a longitudinal study, Psychiatry: Interpersonal and Biological Process-
es, 2015 (in press).
 R. Defrin, K. Ginzburg, M. Mikulincer, Z. Solomon, The long term impact of tissue in-
jury on pain proce ssing and modulation: a study on ex-prisoners of war who
underwent torture, Eur. J. Pain 18 (2013) 548–558.
 K. Prip, A.L. Persson, B.H. Sjölund, Pain when walking: individual sensory proﬁles in
the foot soles of torture victims — a controlled study using quantitative sensory test-
ing, BMC Int. Health Hum. Rights 12 (2012) 40.
 M. Mikulincer, P.R. Shaver, Attachment in Adulthood: Structure, Dynamics, and
Change, Guilford, New York, 2007.
 K. Bartholomew, L.M. Horowitz, Attachment styles among young adults: a test of a
four-category model, J. Pers. Soc. Psychol. 61 (1991) 226–244.
 J. Bowlby, Attachment and Loss: Volume 3. Loss, Sadness and Depression, Basic
Books, New York, 1980.
 J.A. Crowell, D. Treboux, E. Waters, Stability of attachment representations: the tran-
sition to marriage, Dev. Psychol. 38 (4) (2002) 467 (2002).
 J. Davila, R. Cobb, Predicting change in self-reported and interviewer assessed adult
attachment: tests of the individual difference and life stress models of attachment
change, Personal. Soc. Psychol. Bull. 29 (2003) 859–870.
 M. Mikulincer, T. Ein-Dor, Z. Solomon, P.R. Shaver, Trajectories of attachment insecu-
rities over a 17-year period: a latent growth curve analysis of the impact of war cap-
tivity and posttraumatic stress disorder, J. Soc. Clin. Psychol. 30 (2011) 960–984.
 Z. Solomon, R. Dekel, M. Mikulincer, Complex trauma of war captivity: a prospective
study of attachment and posttraumatic stress disorder, Psychol. Med. 38 (2008)
 T.J. Sharp, A.G. Harvey, Chronic pain and posttraumatic stress disorder: mu tual
maintenance? Clin. Psychol. Rev. 21 (2001) 857–877.
 A. Kraus, E. Geuze, C. Schmal, W. Greffrath, R. Treede, M. Bohus, E. Vermetten, Differ-
entiation of pain ratings in combat-related posttraumatic stress disorder, Pain 143
 J.D. Otis, T.M. Keane, R.D. Kerns, An examination of the relationship between chronic
pain and post-traumatic stress disorder, J. Rehabil. Res. Dev. 40 (2003) 397–406.
 R. Stam, PTSD and stress sensitisation: a tale of brain and body part 1: human stud-
ies, Neurosci. Biobehav. Rev. 31 (2007) 530–557.
 R. Defrin, K. Ginzburg, Z. Solomon, E. Polad, M. Bloch, M. Govezensky, S. Schreiber,
Quantitative testing of pain perception in subjects with PTSD—implications for the
mechanism of the coexistence between PTSD and chronic pain, Pain 138 (2)
 M. Mikulincer, P.R. Shaver, N. Horesh, Attachment bases of emotion regulation and
posttraumatic adjustment, in: D.K. Snyder, J.A. Simpson, J.N. Hughes (Eds.), Emotion
Regulation in Couples and Families: Pathways to Dysfunction and Health, American
Psychological Association, Washington 2006, pp. 77–99.
 G. Emilien, C. Penasse, G . Charles, D. Martin, L. Lasseaux, A. Waltregny, P ost-
traumatic stress disorder: hypotheses from clinical neuropsychology and psycho-
pharmacology research, Int. J. Psychiatry Clin. Pract. 4 (2000) 3–18.
 P.J. Meredith, T. Ownsworth, J. Strong, A review of the evidence linking adult attach-
ment theory and chronic pain: presenting a conceptual model, Clin. Psychol. Rev. 28
 T.E. Andersen, O. Brink, A. Elklit, PTSD symptoms mediate the effect of attachment
on pain and somatisation after whiplash injury, Clin. Pract. Epidemiol. Ment. Health
2] N.E. Andrews, P.J. Meredith, J. Strong, Adult attachment and reports of pain in
experimentally-induced pain, Eur. J. Pain 15 (2011) 523–530.
 P.J. Meredith, J. Strong, A. Feeney, The relationship of adult attachment to emotion,
catastrophizing, control, threshold and tolerance, in experimentally-induced pain,
Pain 120 (2006) 44–52.
 J.M. Meredith, A review of the evidence regarding associations between attachment
theory and experimentally induced pain, Curr. Pain Headache Rep. 17 (2013) 326.
 C.F. Sambo, M. Howard, M. Kopelman, S. Williams, A. Fotopoulou, Knowing you care:
effects of perceived empathy and attachment style on pain perception, Pain 151
 C.L. Wilson, M.A. Ruben, A pain in her arm: attachment, social support, and the tour-
niquet task, Pers. Relat. 18 (2011) 241–265.
 N.I. Eisenberger, S.L. Master, T.K. Inagaki, S.E. Taylor, D. Shirinyan, M.D. Lieberman,
B.D. Naliboff, Attachment ﬁgures activate safety signal-related neural region and re-
duce pain experience, PNAS 108 (28) (2011) 11721–11726.
 A.C. Rowe, K.B. Carnelley, J. Harwood, D. Micklerwright, L. Russouw, C.L. Rennie, C.
Liossi, The effect of attachment orientation priming on pain sensitivity in pain-free
individuals, J. Soc. Pers. Relat. 29 (2012) 488–507.
 Z. Solomon, D. Horesh, T. Ein-Dor, A. Ohry, Predictors of PTSD trajectories following
captivity: a 35-year longitudinal study, Psychiatry Res. 199 (3) (2012) 188–194.
 G. Zeilig, S. Enosh, D. Rubin-Asher, B. Lehr, R. Defrin, The nature and course of sen-
sory changes following spinal cord injury: predictive properties and implications
on the mechanism of central pain, Brain 135 (2012) 418–430.
 R. Melzack, The McGill pain questionnaire: major properties and scoring methods,
Pain 1 (1975) 1277–1299.
 R.C. Fraley, K.E. Davis, Attachment formation and transfer in young adults' close
friendships and romantic relationships, Pers. Relat. 4 (1997) 131–144.
 K.A. Brennan, C.L. Clark, P.R. Shaver, Self-report measurement of adult romantic at-
tachment: an integrative overview, in: J.A. Simpson, W.S. Rholes (Eds.), Attachment
Theory and Close Relationships), Guilford Press, New York 1998, pp. 46–76.
 Z. Solomon, R. Benbenishty, Y. Neria, M. Abramowitz, K. Ginzburg, A. Ohry, Assess-
ment of PTSD: validation of the revised PTSD Inventory, Isr. J. Psychiatry Relat. Sci.
30 (1993) 110–115.
 American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disor-
ders, 4th ed. American Psychiatric Association, Washington DC, 1994.
 Z. Solomon, D. Horesh, Changes in diagnostic criteria of PTSD: implications from two
prospective longitudinal studies, Am. J. Orthopsychiatry 77 (2007) 182–188.
 A.F. Hayes, Introduction to Mediation, Moderation, and Conditional Process Analy-
sis: A Regression-based Approach, Guilford Press, New York, 2013.
 A.J. Carinci, P. Mehta, P.J. Christo, Chronic pain in torture victims, Curr. Pain Head-
ache Rep. 14 (2010) 73–79.
 G. MacDonald, M.R. Leary, Why does social exclusion hurt? The relationship be-
tween social and physical pain, Psychol. Bull. 131 (2) (2005) 202–223.
476 T.E. Andersen et al. / Journal of Psychosomatic Research 79 (2015) 471–476