Journal of Psychosomatic Research 58 (2005) 279 – 288
Adult attachment, alexithymia, and symptom reporting
An extension to the four category model of attachment
Alison J. Wearden*, Naomi Lamberton, Nicola Crook, Victoria Walsh
School of Psychological Sciences, University of Manchester, Oxford Road, Manchester, M139PL
Received 18 February 2004; accepted 7 September 2004
Abstract
Objective: A previous study using a three- category attachment
model found that avoidant attachment was associated with
increased symptom reporting, and that this relationship was largely
mediated by negative affectivity and alexithymia. The present
study aimed to advance on these findings by using a four-category
model of attachment to determine which aspect of avoidant
attachment (fearful or dismissing) is related to symptom reporting,
and via which mediating variables. Method: One hundred and
forty-two male and female undergraduates, aged 17–44, completed
questionnaire measures of attachment style, alexithymia, self-
esteem, positive and negative affectivity, and symptom reporting.
Results: Fearful and preoccupied attachment styles, negative
affectivity, and alexithymia were all significantly associated with
increased symptom reporting, while the dismissing attachment
Keywords: Adult attachment; Alexithymia; Symptom reporting; Negative affectivity; Self-esteem
Introduction
Since Hazan and Shaver demonstrated that infant attach-
ment styles, as described in Ainsworth’s classic studies,
persisted into adulthood, the study of adult attachment and
its correlates has been a rapidly growing area of research
[1–3]. These attachment styles, formed in large part as a
result of our childhood experiences with caregivers, are
conceptualised as internal representations or working
models of patterns of relating with others. They are
therefore thought to influence our social interactions
throughout life. Indeed, a 20-year longitudinal study found
* Corresponding author. Department of Psychology, University of
Manchester, Oxford Road, Manchester M13 9PL, UK. Tel.: +44 161 275
2684; fax: +44 161 275 2588.
E-mail address: alison.wearden@manchester.ac.uk (A.J. Wearden).
0022-3999/04/$ – see front matter D 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.jpsychores.2004.09.010
style was not. Regression analyses showed that the relationship
between fearful attachment and symptom reporting was partly, but
not fully, mediated by alexithymia and negative affectivity, while
that between preoccupied attachment and symptom reporting was
mediated mainly by negative affectivity. Low self-esteem was
associated with symptom reporting only via its association with
negative affectivity. Conclusions: Fearful and preoccupied attach-
ment styles are both associated with symptom reporting via a
negative model of the self and increased negative affectivity, but
alexithymia is an additional predictor of symptom reporting in
individuals with fearful attachment. This difference is thought to
be linked to the model of others developed in early interactions
with caregivers.
D 2005 Elsevier Inc. All rights reserved.
that 72% of young adults had the same attachment
classification as they had when they were infants, and that,
where attachment classification had changed, severe, threat-
ening life events were often implicated [4]. Recently,
researchers have started to examine the link between
attachment style and various health-related variables, such
as health care use and, of relevance to this study, symptom
reporting [5–9]. Work in this area may help health
professionals to understand and treat their patients better.
The theoretical basis of attachment, and the best way to
conceptualise it, has been widely discussed [3]. Whereas
originally, three categories of attachment (secure, anxious,
and avoidant) were proposed [1,2], an alternative model
with four categories has recently been developed [10]. The
latter model has three insecure attachment styles: preoccu-
pied, equivalent to the original anxious style, and fearful and
dismissing, which represent subdivisions of the original
280 A.J. Wearden et al. / Journal of Psychosomatic Research 58 (2005) 279–288
avoidant attachment style. The four-category model is based
on the idea that, as a result of early experiences, people
develop lasting internal representations of self in relation-
ships (model of self ) and representations of others in
relationships (model of other), as illustrated in Fig. 1. These
representations influence whether people perceive them-
selves as worthy of care and whether others can be trusted to
provide care [10].
An adult with a typical secure attachment style is thought
to have received consistent, responsive care-giving during
childhood. This person has a positive model of self and
others and is comfortable turning to others for help and
being comforted by others. The preoccupied attachment
style is likely to arise from inconsistently responsive early
care-giving, thus, this style is characterized by poor self-
esteem (negative model of self ) and focus on negative
affect. The individual with a preoccupied attachment style is
likely to have an idealised model of others, to be very needy
of others, and to frequently seek reassurance. A predom-
inantly fearfully attached individual shares the negative
view of self with the preoccupied person, feels a need for
social relationships, but holds a negative view of others.
This style of relating is thought to be due to harsh or
rejecting care-giving, leading to a fear of intimacy fuelled
by fear of rejection. Finally, the dismissing attachment style
is believed to be related to consistently unresponsive early
care-giving, which led the individual to become compul-
sively self-reliant due to their negative view of others and
positive view of self. While adults can be categorised into
the four attachment styles (either by self- categorisation or
using questionnaire scores), scale measures of the extent to
which people resemble the different attachment styles are
also used, as in the present study.
Three studies using undergraduate students as partici-
pants have found a link between insecure attachment style
and symptom reporting. Our previous study, based on the
three-category model of attachment, and using the Adult
Attachment Scale [11], found that avoidant attachment was
Model of self
Positive Negative
Secure Preoccupied
Positive Trusts others and Idealises others,
feels worthy of emotionally needy,
others’ attention seeks reassurance
Model
of
other Dismissing Fearful
High-self-worth, Approach-
Negative
compulsively self- avoidance, fears
reliant intimacy
Fig. 1. The four-category model of attachment. (Adapted from Bartholo-
mew and Horowitz [10]).
weakly predictive of symptom reporting and emotional
preoccupation as a way of coping with health problems [9].
A similar study using the same three-category model also
linked avoidant attachment style to increased symptom
reporting and high levels of emotional control or suppres-
sion [7]. However another study reported that it was the
anxious attachment style that was linked to higher levels of
symptom reporting, but that the relationship was largely
mediated by negative affectivity [6].
It is possible that the inconsistent results described above
arise from the use of different measures of adult attachment.
More importantly, it may be that the association between
symptom reporting and both anxious and avoidant attach-
ment in different studies is due to some overlap between these
categories. As noted above, the four-category model of
attachment divides the avoidant style into fearful and
dismissing attachment, while anxious attachment style is
most like the preoccupied style. A recent study used both
categorical and scale measures to assess the attachment styles
of female medical patients in accordance with the four-
category model and found that participants categorised as
having preoccupied or fearful attachment (i.e., those with a
negative view of themselves) reported a significantly greater
number of physical symptoms than securely attached patients
[5]. These authors’ findings using continuous measures of
attachment were consistent with those from the categorical
analysis, with a strong positive correlation between fearful
attachment and somatic symptom reporting. It is therefore
possible that, in previous work using a three-category model,
it was the fearful aspect of avoidant attachment, which shares
features in common with preoccupied attachment, that was
associated with higher symptom reporting.
Why should there be a link between symptom reporting
and attachment style? Negative affectivity, or the general
tendency to experience and communicate negative emo-
tions, has repeatedly been shown to predict the extent to
which people report physical symptoms, possibly via the
mechanism of enhanced attention to and perception of
bodily sensations [12]. It has been shown that negative
affectivity can act as an important mediator between
attachment style and symptom reporting [6,9]. It has been
proposed that the link between preoccupied and fearful
attachment and symptom reporting might be based on the
low self-esteem and tendency to focus on negative affect,
which arises from having a negative model of self [5]. To
date, however, no study examining the association between
the four-category model of attachment and symptom
reporting has tested this proposition by including specific
measures of self-esteem and negative affectivity.
A second variable that has repeatedly been shown to
correlate with symptom reporting and that has been
implicated as a mediator between attachment style and
symptom reporting is alexithymia [13]. Alexithymia is
conceptualised as a deficit in the ability to identify and
describe emotions, but not to experience them, coupled with
a tendency to externally oriented and concrete thinking [14].
 A.J. Wearden et al. / Journal of Psychosomatic Research 58 (2005) 279–288
It is thought that what we come to know about our emotions
and our ability to understand, describe, and regulate our
affect are strongly influenced by early interactions with
caregivers, much in the same way as attachment style.
Experience of caregivers who do not display or discuss their
emotions, or who do not deal appropriately with the child’s
emotions, is thought to have a major influence on affect
regulation later in life [15]. It has been suggested that, after
the age of about 3 months, infants are particularly biased to
perceive exteroceptive cues, among them the facial and
vocal responses of caregivers. These cues are thought to
play a particularly important role in regulating infant
distress, and from them, infants learn to be aware of, and
ultimately to regulate, their own emotions [16].
Nowadays, alexithymia is usually measured using the
self-report Toronto Alexithymia Scale (TAS-20; [17]) and is
most commonly conceptualised as a continuum, although
some studies categorise people as alexithymic or not using a
cut-off on the TAS-20 [18]. It was suggested some time ago
that people who are unable to identify their emotions as such
might misattribute the somatic manifestations of emotions
as physical symptoms and, thus, that alexithymia might
provide a partial explanation for the presence of medically
unexplained symptoms or somatization [19]. Commentators
agree that, while receiving some support in the literature,
this suggestion has not been adequately tested [20,21].
However, whether or not alexithymia places a person at
greater risk for medically unexplained as opposed to
medically explained symptoms, it has been established that
certain aspects of alexithymia, particularly difficulty iden-
tifying feelings, are associated with increased symptom
reporting [21]. It has previously been shown, using a three-
category model of attachment, that both negative affectivity
and alexithymia are significant mediators of the association
between avoidant attachment and symptom reporting [9].
Furthermore, a recent study of young men with mood
symptoms found that alexithymic traits were more pro-
nounced in individuals who had insecure patterns of
attachment [22]. Using the four-category model of attach-
ment, those with preoccupied and fearful styles had a higher
prevalence of alexithymia than did participants with a
dismissing pattern. Thus, there is some evidence for the idea
that disordered relationships in infancy, leading to insecure
attachment, are associated with experiencing more negative
emotion and a decreased ability to understand that emotion,
and that these factors predispose a person to experience
somatic sensations as symptoms.
The present study was designed to replicate and extend
the findings of previous work by examining the potential
mediating roles of self-esteem, negative affectivity, and
alexithymia in the association between attachment style and
symptom reporting, but this time, using the four-category
model of attachment, to allow the separation of the avoidant
style into fearful and dismissive styles. In an attempt to
clarify the findings from previous work [9], we hypothe-
sized that the dismissing avoidant attachment style, which
281
entails a positive model of the self, would not be correlated
with negative affect or low self-esteem. To the extent that
the link between symptom reporting and attachment style is
mediated by negative affectivity, we would not therefore
expect dismissive attachment to be correlated with symptom
reporting. On the other hand, we expected to replicate, but
this time in a nonclinical as opposed to a medical
population, the finding that continuous measures of avoi-
dant fearful and preoccupied attachment are positively
correlated with symptom reporting [5] and to extend this
work by exploring the mediators of the attachment style–
symptom reporting link. To this end, we hypothesized that
the preoccupied and fearful attachment styles, both of which
involve a negative model of the self, would be related to
lower self-esteem and increased negative affectivity, and we
wished to determine whether low self-esteem was a
predictor of symptom reporting, independent of its associ-
ation with negative affectivity. Finally, we hypothesized that
preoccupied and fearful attachment would be associated
with higher levels of alexithymia, and we expected that both
alexithymia and negative affectivity would mediate (at least
partly) the association between the preoccupied and fearful
attachment styles and symptom reporting.
Method
This study was approved by the Ethics Committee of the
University of Manchester Department of Psychology, in
accordance with guidelines provided by the British Psycho-
logical Society.
Participants
First year undergraduate students in the Department of
Psychology were invited to participate in this study in
exchange for course credit. Participants were recruited via
advertisements in the department and announcements in
lectures and were free to choose whether to participate in
this or other studies on offer. One hundred and forty-four of
195 first year students (74%) chose to take part in this study.
They were aged from 17 to 44 years (mean age = 19.42,
S.D. =3.34). One hundred and twenty-three participants
(86.6%) were female. One hundred and twenty-three
participants (86.6%) gave their ethnicity as white, 7 (4.9%)
as Indian or Pakistani, and the remaining 12 participants
(8.5%) were of other ethnic origins. The gender and ethnic
breakdown of the participants reflected that of the popula-
tion as a whole. Information on marital status was
not available.
Procedure
Participants who expressed an interest in the study were
given an information sheet to read, a consent form to sign,
and a questionnaire booklet containing the measures
282 A.J. Wearden et al. / Journal of Psychosomatic Research 58 (2005) 279–288
described below. Participants completed the questionnaires
individually and confidentially, and after the course credit
had been allocated, the questionnaires were anonymised for
data entry. Data from two participants were discarded
because of missing answers.
Measures
All booklets contained the questionnaires in the order in
which they are described below.
A demographic questionnaire asked for age, sex, and
ethnic classification.
Attachment was measured using the Relationship Ques-
tionnaire [3], which we chose partly because of its ease of
administration but also because it had been used in the study
by Ciechanowski et al. [5], which we were hoping to
replicate in a nonclinical population. The Relationship
Questionnaire consists of four short descriptive paragraphs,
one for each of the four attachment styles. In this study,
participants rated on a seven-point scale ranging from not at
all like me to very much like me the extent to which each
description was like them. The best-fitting attachment style
was determined categorically for each subject based on the
attachment category with the highest score. It was also
possible to derive measures of the underlying schemas,
model of self, and model of other, using the scores on the
attachment style scales, as follows:
model of self ¼ ðsecure m dismissingÞ
 ðpreoccupied m fearfulÞ;
model of other ¼ ðsecure m preoccupiedÞ
 ðdismissing m fearfulÞ:
Self-esteem was measured using the 10 item Rosenberg
Self-esteem Questionnaire [23]. Scores on this scale could
range from 10 to 50, with a higher score indicating higher
self-esteem. In this study, the scale had an a coefficient of .88.
Alexithymia was measured using the TAS-20 [17]. This
is a 20-item scale that assesses three aspects of the
alexithymia construct: difficulty identifying feelings (seven
items), difficulty describing feelings (five items), and
externally orientated thinking (eight items). In the present
study, the internal consistency (a) coefficients for the three
scales were .84 (DIF ), .80 ( DDF ) and .61 (EOT ). The poor
internal reliability and questionable validity of the EOT
scale has been noted previously [24]. It has also been noted
that the difficulty identifying feelings and difficulty describ-
ing feelings scales are actually tapping similar constructs,
neither of which correlate very highly with the externally
oriented thinking scale [25,26]. Given these findings, the
first two scales were combined to produce a new scale,
difficulty identifying and describing feelings, with an a
coefficient of .87, and this scale was used in the main
analyses. Scores on this new scale could range from 12 to
60, with higher scores indicating more alexithymia. To aid
comparison with other work, all analyses in this study were
repeated using the entire TAS-20 scale, and the results of
these analyses are reported in the text.
Positive and Negative Affectivity were measured using
the PANAS [27]. This instrument consists of 10 positive and
10 negative adjectives and can be used to measure trait or
state affect. Participants rated each adjective on a scale of 1
(very slightly or not at all) to 5 (extremely) for the extent to
which they usually felt that way, thus providing measures of
trait positive and negative affectivity. Responses were
summed to give positive and negative affectivity scores
each with a range of 10 to 50. In this study, positive and
negative affectivity scales were significantly inversely
correlated (r =.253, P b.005), and the internal consistency
of each scale was high (a = .87 and .88, respectively).
Symptom reporting was measured using a 20-item scale
asking participants to report the frequency of common
minor symptoms or sensations on a five-point scale from
have never or almost never experienced to experience more
than once every week. This response scale is similar to that
used in the PILL symptom reporting questionnaire [28].
Thirteen of the symptoms were those used in the Symptom
Interpretation Questionnaire [29], plus seven additional
symptoms or sensations. The complete list of symptoms
inquired about was as follows: sore throat, headache, feeling
dizzy, tight feeling in chest, feeling hot and sweaty, hands
shaking and trembling, difficulty sleeping, feeling cold and
shivery, bbutterflies Q in tummy, nausea (feeling sick),
numbness or tingling in hands or feet, heart pounding or
racing, pain in muscles, difficulty swallowing, feeling
exhausted, dry mouth, weakness in limbs, difficulty
concentrating, feeling faint, and gone off certain food,
coffee or alcohol. The scores for each symptom were
summed to give a total symptom frequency score ranging
from 20 to 100, with an a coefficient of .88.
The last questionnaire in the booklet asked participants to
give their own causal explanations for the symptoms listed
Table 1
Mean (S.D.) scores for the total sample (N=142) on the attachment style,
alexithymia, positive and negative affectivity, self-esteem, and symptom
reporting measures
Scale (possible range of scores) Mean S.D.
Secure attachment style score (1–7) 4.51 1.43
Anxious-preoccupied attachment style score (1–7) 3.58 1.69
Avoidant-fearful attachment style (1–7) 3.56 1.88
Avoidant-dismissing attachment style (1–7) 3.24 1.60
Rosenberg self-esteem scale (10–50) 36.56 6.45
Alexithymia: diff iculty identifying and 28.16 7.65
describing feelings (12–60)
Alexithymia: externally oriented thinking (8–40) 18.46 3.77
Alexithymia: TAS-20 total score (20–100) 46.62 9.14
PANAS positive affectivity (10–50) 33.71 6.15
PANAS negative affectivity 10–50) 22.42 7.21
Symptom scale total (20–100) 50.53 11.46
 A.J. Wearden et al. / Journal of Psychosomatic Research 58 (2005) 279–288 283

Table 2
Mean (S.D.) symptom reporting score of participants categorised into each
of the four attachment styles
Attachment style Mean S.D. F(3,129) Post hoc
Secure 48.0 11.0 3.719 Fearful N dismissing
Dismissing 45.7 12.6 P = .013
Preoccupied 53.1 10.1
Fearful 54.9 12.1
above and finally to state the number of times they had been
to see a general practitioner (GP) in the past 6 months. The
findings from the symptom attribution questionnaire will not
be reported further here.
Statistical analysis
Data were analysed using SPSS 10.1 for Windows. The
distributions of the main variables to be used in the analysis
were screened for normality. One way analysis of variance
(ANOVA) was used to examine differences in symptom
reporting between the participants categorised in the four
attachment styles, with post hoc analysis using the Student
Newman Keul’s test. Pearson’s correlations were computed
among all four attachment style scores: alexithymia self-
esteem positive and negative affectivity variables and self-
esteem (collectively called the predictor variables) on the
one hand and the outcome variable, the symptom reporting
measure, on the other. Data from all 142 participants were
used in the analysis. Hierarchical multiple linear regressions
were then carried out to explore whether the predictor
variables were each independently associated with the
outcomes. Finally, the relationship between model of other,
model of self, and the other variables was explored. The
regression models were checked for linearity by inspection
of residual plots. Possible multicollinearity was checked
using the tolerance and variance inflation factor (VIF )
statistics. In all cases, the tolerance value was .8 or above,
and the VIF statistic was 1.281 or below, confirming that
multicollinearity was not a problem in this data set.
Results
Table 1 shows the mean scores obtained in the whole
sample on the various questionnaire measures.
In nine cases, it was not possible to allocate participants
to an attachment style because they endorsed two or more
attachment style scales equally. Of the remaining 133
participants, the percentages categorised in each attachment
style were as follows: secure (n =68, 51.1%), preoccupied
(n =24, 18.0%), fearful (n =29, 21.8%), and dismissing
(n =12, 9.0%).
Sixty-eight (47.9%) of the participants reported having
visited their GP at least once in the specified period; of
these, 41 (29% of total) had visited the GP once, 16 (11% of
total) twice, 7 (5% of total) three times, and 4 (3% of total)
four times. A cross-tabulation of attachment style versus
having visited the GP or not revealed no significant
association between these two variables [v 2(3 df )= 2.569,
P =.463]. There was no significant difference on any of the
measures of alexithymia, affectivity, self-esteem, or symp-
tom reporting between participants who reported visiting
their GP at least once and those who did not visit their GP.
Spearman’s correlations between the number of GP visits
and the aforementioned measures were all of .140 or less,
and none was statistically significant.
A one-way ANOVA comparing total symptom reporting
scores for the 133 participants categorised in the four
attachment styles revealed a significant overall effect of
attachment style, with post hoc analyses showing that
participants classified as having fearful attachment reported
significantly more symptoms than did those classified as
having dismissing attachment (see Table 2).
Pearson’s correlations between the predictor variables
(attachment style scores, alexithymia, positive and negative
affectivity, self-esteem, age, and sex) and symptom report-
ing were computed. No significant effects of age or sex were
found. Positive affectivity was significantly correlated with
self-esteem (r = .571, P b.001) and weakly with secure
attachment (r = .181, P = .030). Positive affectivity was
significantly, negatively correlated with alexithymia

Table 3
Pearson’s correlations between predictor variables (attachment style scores, negative affectivity, and alexithymia) and outcome variable, total symptom
reporting score
Avoidant Avoidant Negative Symptom
Anxious preoccupied fearful dismissing affectivity Alexithymiaa Self-esteem reporting
Secure .178* .407*** .134 .364*** .348*** .387*** .271***
Anxious preoccupied – .284*** .214** .269*** .196* .312*** .250***
Avoidant fearful – .247*** .296*** .337*** .314*** .337***
Avoidant dismissing – .059 .120 .028 .038
Negative Affectivity – .458*** .569*** .497***
Alexithymiaa – .372*** .467***
Self-esteem – .258**
a
Difficulty identifying and describing feelings.
* P b.05.
** P b.01.
*** P b.005.
284 A.J. Wearden et al. / Journal of Psychosomatic Research 58 (2005) 279–288

Table 4 predictor variables showed that, when self-esteem alone was

Regression equation predicting symptom reporting on the basis of alexithy-
mia (difficulty identifying and describing feelings) and negative affectivity
Variables B S.E. B b t
a
Alexithymia 0.454 0.118 .303 3.854
Negative affectivity 0.569 0.125 .358 4.551
Constant 24.974 3.316 7.531
Final adjusted R 2 = .310, F(2,139) for regression = 32.661, P b.001
a
Difficulty identifying and describing feelings.
(r = .280, P =.001) and the alexithymia bexternally ori-
ented thinking Q subscale (r = .262, P = .002) and weakly
with symptom reporting (r =.186, P =.026). b Externally
oriented thinking Q (EOT ) was not significantly correlated
with any other variable, the highest correlation being
between EOT and dismissive attachment (r = .145,
P = .086). All other correlations are shown in Table 3.
As can be seen in Table 3, both preoccupied and fearful
attachment styles were positively associated with symptom
reporting, and secure attachment was negatively associated
with symptom reporting. Secure attachment was associated
with higher self-esteem, lower negative affectivity, and
lower alexithymia, while preoccupied and fearful attach-
ment were both associated with lower self-esteem and
greater negative affectivity. Avoidant-dismissing attachment
was unrelated at the predetermined significance level of
P = .005 with any variable other than avoidant-fearful
attachment. Only fearful attachment was associated with
alexithymia at the predetermined significance level. Addi-
tionally, lower self-esteem, higher negative affectivity, and
more alexithymia were all associated with increased
symptom reporting and with each other.
The first regression analyses explored whether low self-
esteem predicted symptom reporting, independent of its
relationship with negative affectivity. An hierarchical multi-
ple regression analysis with symptom reporting as the
outcome and self-esteem and negative affectivity as the
in the equation, it predicted 6.6% of the variance in
symptom reporting [R 2 = .066, F(1,140) = 9.952, P = .002].
Significance
When negative affectivity entered the equation with self-
P b.001 esteem already in, it accounted for an additional 18.2% of
P b.001
the variance in symptom reporting [R 2 change = .182, F for
P b.001
change (1,139) =33.565, Pb.001]. On the entry of negative
affectivity into the equation, the standardised b weight for
self-esteem reduced from .258 (t =3.155, P =.002) to
.037 (t=.418, P =.677), showing that the effect of self-
esteem on symptom reporting was almost entirely mediated
by negative affectivity.
The next set of analyses explored the associations
between attachment style, negative affectivity, alexithymia,
and symptom reporting. An hierarchical multiple regression
analysis with symptom reporting as the outcome and
alexithymia and negative affectivity as the predictor
variables showed that, when alexithymia alone was in the
equation, it predicted 21.8% of the variance in symptom
reporting [R 2 = .218, F(1,140) = 39.106, P b.001]. When
negative affectivity entered the equation with alexithymia
already in, it accounted for an additional 10.1% of the
variance in symptom reporting [R 2 change = .101,
F(1,139) =32.661, P b.001], and alexithymia remained a
significant predictor (standardised b value reduced from
.467, t = 6.254, P b.001, to .303, t =3.854, P b.001). There-
fore, alexithymia and negative affectivity were both found to
be significant independent predictors of symptom reporting
(see Table 4). When these analyses were repeated, substitut-
ing the TAS-20 total score as the alexithymia variable, the
final b weight for TAS-20 was .219 ( P =.006), and the
significance levels of the other variables was not altered.
Next, three separate hierarchical regression analyses were
carried out, in which attachment style (fearful, preoccupied,
or secure) was entered on the first step, negative affectivity
(the strongest single predictor of symptom reporting) on the
second step, and alexithymia on the third. In each case,

Table 5
Final regression equations predicting total symptom reporting score on the basis of alexithymia (difficulty identifying and describing feelings), negative
affectivity and (a) fearful, (b) preoccupied, and (c) secure attachment styles
Variables B S.E. B b t Significance
(a) Alexithymiaa 0.397 0.120 .265 3.307 P = .001
Neg. affect 0.526 0.125 .331 4.196 P b.001
Fearful 0.911 0.454 .150 2.006 P = .047
Constant 24.298 3.298 7.368 P b.001
Final adjusted R 2 = .325, F(3,138) for regression = 23.589, P b.001
(b) Alexithymiaa 0.440 0.118 .294 3.733 P b.001
Neg. affect 0.532 0.127 .335 4.180 P b.001
Preoccupied 0.693 0.493 .102 1.408 P = .161
Constant 23.718 3.423 6.929 P b.001
Final adjusted R 2 = .315, F(3,138) for regression = 22.588, P b.001
(c) Alexithymiaa 0.440 0.121 .294 3.630 Pb.001
Neg. affect 0.551 0.129 .347 4.265 Pb.001
Secure 0.341 0.619 .042 .551 P =.583
Constant 27.314 5.395 5.063 P b.001
Final adjusted R 2 = .306, F(3,138) for regression = 21.766, P b.001
a
Difficulty identifying and describing feelings.
 A.J. Wearden et al. / Journal of Psychosomatic Research 58 (2005) 279–288
Table 6
Pearson’s correlations between model of self, model of other and negative
affectivity, self-esteem, alexithymia, and symptom reporting
Negative Self-
affectivity esteem Alexithymia TAS-20
Model .372*** .444*** .329*** .236**
of self
Model .193* .154 .267*** .254**
of other
* P b.05.
** P b.005.
*** P b.001.
attachment style was a significant predictor of symptom
reporting when it alone was in the equation. In the case of
fearful attachment, attachment style remained a significant
predictor even after the entry of negative affectivity and
alexithymia (see Table 5 for final equation), thus suggesting
that the association between fearful attachment and symp-
tom reporting was partially, but not fully, mediated by both
negative affectivity and alexithymia. In the case of
preoccupied attachment, the entry of negative affectivity
into the equation on the second step reduced the stand-
ardised b weight for preoccupied attachment from .250
(t = 3.056, P =.003) to .125 (t = 1.657, P =.100), suggesting
that negative affectivity was an important mediating variable
in the association between preoccupied attachment and
symptom reporting. Finally, a similar analysis, starting off
with secure attachment in the equation, revealed that the
negative zero-order correlation between secure attachment
and symptom reporting was mediated both by negative
affectivity (standardised b for secure attachment reduced
from .271, t = 3.330, P = .001, to .104, t = 1.322,
P = .188, when negative affectivity entered the equation)
and by alexithymia (b further reduced to .042, t = .551,
P = .583, when alexithymia entered on the third step). In this
case, it appears that the negative correlation between secure
attachment and symptom reporting is almost entirely due to
the negative correlations between secure attachment, neg-
ative affectivity, and alexithymia. The parameters for the
final equations are shown in Table 5.
When all of the above analyses were repeated substitu-
ting TAS-20 as the alexithymia variable, similar patterns of
results were obtained. TAS-20 proved to be a significant pre-
dictor of symptom reporting in each case, and the signifi-
cance of the contribution of the other variables was unaltered.
Table 7
Multiple regression equation predicting total symptom reporting on the basis of negative affectivity, alexithymia (difficulty identifying and describing feelings),
and model of self
Variables B S.E. B
Alexithymiaa 0.397
Negative affectivity 0.485
Model of Self 0.560
Constant 28.863
Final adjusted R 2 = .334, F(3,138) for regression = 24.755, P b.001
a
Difficulty identifying and describing feelings.
285
Turning next to the measures of the models of self and
other that are thought to underlie attachment style, Pearson’s
correlations showed that the other and self models were not
Symptom
correlated with each other (r =.080, P =.344), suggesting
reporting
that these are indeed orthogonal dimensions of internal
.389***
representations. As can be seen in Table 6, both models
.145 were significantly, negatively correlated with alexithymia,
but only the model of self was significantly, negatively
correlated with negative affectivity and symptom reporting
and was significantly, positively correlated with self-esteem.
Multiple regression analysis in which negative affectivity
and alexithymia were entered on the first step and model of
self on the second showed that the model of self was a
significant predictor of symptom reporting, even when
negative affectivity and alexithymia were already in the
equation (see Table 7 for parameters for final equation).
Thus, the power of the model of self to predict symptom
reporting was partially mediated by alexithymia and
negative affectivity, but the model of self also made a
significant unique contribution to the equation. Once again,
when the analysis was repeated substituting the TAS-20
total score as the alexithymia variable, a similar pattern of
results was obtained, although TAS-20 proved to be a
somewhat weaker predictor (final b weight =.198, P = .012).
Discussion
As expected, in this study, we found that there were
significant, positive associations between scores on scales
measuring preoccupied and fearful attachment and symptom
reporting. We therefore replicated in a young, mixed-sex,
nonclinical population what had previously been found in a
female general medical population [5]. Furthermore, we
obtained a significant, negative correlation between the
score for secure attachment and symptom reporting and
found no association between the dismissing attachment
style scale score and symptom reporting. In view of the
theoretical suggestion that both fearful and preoccupied
attachment styles involve a negative model of the self, we
expected that both would be associated with lower self-
esteem and higher levels of negative affect, and this is what
we found. Additionally, both preoccupied and fearful
attachment styles were associated with higher levels of
alexithymia, although the correlation between preoccupied
b t Significance
0.118 .265 3.370 P = .001
0.127 .305 3.813 P b.001
0.224 .189 2.504 P = .013
3.603 8.010 P b.001
286 A.J. Wearden et al. / Journal of Psychosomatic Research 58 (2005) 279–288
attachment style and alexithymia was weak. Regression
analyses showed that the relationship between fearful
attachment and symptom reporting was partially mediated
by alexithymia and negative affectivity, each of which
contributed independently to symptom reporting. The
relationship between preoccupied attachment and symptom
reporting was largely mediated by negative affectivity. Self-
esteem mediated the relationship between fearful and
preoccupied attachment and symptom reporting, but not
independently of its association with negative affectivity.
Further analysis suggested that the association between a
negative model of the self, which underlies preoccupied and
fearful attachment, and symptom reporting was partly
mediated by both negative affectivity and alexithymia.
Our previous study, using the Hazan and Shaver [1]
three-category model of attachment linked the avoidant
attachment style to symptom reporting, but the link was
weak. The present study aimed to clarify that association by
using the four-category model of attachment. In the four-
category model, avoidant attachment is represented by the
dismissing and fearful styles. The present study found the
fearful attachment style to be a predictor of symptom
reporting (even when alexithymia and negative affectivity
were controlled for), but dismissing attachment was not.
Thus, the findings of our earlier work [9] were replicated but
clarified by the finding that it is only the bfearfulQ and not
the bdismissing Q aspect of avoidant attachment that is
associated with increased symptom reporting.
Our results are consistent with the finding that medical
patients with preoccupied and fearful attachment and,
therefore, a negative model of self reported a significantly
greater number of physical symptoms than did securely
attached patients [5]. The authors of that study proposed that
b symptom reporting may be positively associated with
one’s self-esteem and tendency to experience distress Q
(p. 665). We tested this proposition, and our findings
suggest that, although low self-esteem is indeed related to
fearful and preoccupied attachment and is moderately
strongly correlated with a negative model of the self, its
role as a predictor of symptom reporting is almost entirely
due to its association with negative affectivity.
In the present study, both models of self and of other
were negatively associated with alexithymia, thus, scores on
the secure attachment scale (which incorporates positive
models of both self and other) were negatively correlated
with alexithymia, whereas scores on the fearful attachment
scale (which incorporates negative models of both self and
other) were positively correlated with alexithymia. The
pattern of findings supports the proposition that both
negative affectivity and alexithymia contribute additively
towards symptom reporting [25], and this may explain why
fearful attachment, which involves both high negative affect
and high alexithymia, is most strongly correlated with
symptom reporting. It is of interest that alexithymia is
associated with both negative models of the self and of
others. In terms of attachment theory, fearfully attached
individuals who may have experienced overly harsh or
neglecting care-giving in their early years may learn to deny
or suppress their negative emotions about distressing events,
so as not to risk further distancing or alienation from the
caregiver [30]. Furthermore, it has been reported that higher
levels of alexithymia are associated with diminished family
expressiveness and with feeling less emotionally safe during
childhood [31]. Possibly, people with a negative model of
both themselves and others do not have the opportunity
either to practice or to observe safe emotional expression
and end up with a diminished ability to understand
emotions. Anxious-preoccupied children, on the other hand,
have a positive model of others, tend to be overly vigilant to
negative events, and to express their feelings about such
events strongly, to ensure a response from inconsistent
caregivers [30]. Such children may therefore have more
opportunity to reflect on their emotions, perhaps leading to a
lower level of alexithymia than is seen in the fearful
attachment style.
An alternative theoretical position that helps to explain
the associations between attachment, alexithymia, and
symptom reporting derives from a body of work on
b affect-mirroring Q [16]. This work suggests that the
responses which carers make when infants are distressed
play a crucial role in regulating the infants’ emotions.
Successful affect regulation takes place when carers
appropriately reflect back infants’ emotional states. Infants
recognise the contingency between the regulation of their
affect and the cues that they are receiving from their carers
and therefore come to learn about their emotions through
this interaction. Furthermore, recent work has begun to
measure the physiological correlates of affect– regulation
interactions [32] and suggests a mechanism by which
insecurely attached and highly alexithymic individuals
might experience more somatic sensations.
The dismissing attachment style was not found to be
significantly associated with any other variables. This
finding could be related to the compulsive self-reliance that
characterizes the dismissing style. Dismissing individuals
may be prone to a social desirability bias in that they do not
want to appear weak, or in need, by reporting symptoms or
reporting negative feelings. However, only 9% of our
participants had a dismissing attachment style, whereas we
might have expected around twice this proportion [10], and
this small proportion of dismissing participants hampers the
conclusions that we can draw.
The present study differed from our previous work
principally in the use of the four-category attachment model
rather than the three-category model. Debate continues as to
the best model of attachment, and a recent study has
described bfive latent classesQ of attachment [8]. These
authors also found that attachment style was associated with
symptom reporting and extended the model by finding that
certain attachment styles were more prone to certain types of
symptoms (e.g., fatigue or pain). This provides an interest-
ing possibility for further study.
 A.J. Wearden et al. / Journal of Psychosomatic Research 58 (2005) 279–288
While our findings, with respect to symptom reporting,
were largely as hypothesised, we acknowledge that report-
ing on common somatic symptoms at the request of
researchers may be quite different from the spontaneous
reporting of symptoms to family members or to a medical
practitioner, for example. Our symptom reporting measure
was not related to reported GP visits in the previous 6
months, nor were attachment style measures related to
reported GP visits. In a much larger sample of adult female
primary care patients, it was found that participants with a
fearful attachment style, despite reporting the most somatic
symptoms, actually had lower health-care utilization costs
than did participants with the other attachment styles [5],
suggesting that attachment style contributes differentially to
symptom perception on the one hand and to engaging the
health care system to deal with symptoms on the other. Our
findings provide further evidence for the role of attachment
style in symptom perception.
The present study was restricted to young psychology
undergraduates who consented to participate and who might
be expected to be higher functioning and generally healthier
than a more mixed population. We did not assess the
physical or mental health of the population and were
therefore unable to take account of factors such as medical
comorbidity, depression, and anxiety, the latter two of which
are associated with both adult attachment style [33] and
symptom reporting [34]. Therefore, the results of this study
may not be generalisable to other populations. On the other
hand, the fact that the findings were broadly similar to those
obtained from a general medical population [5] suggests that
the association between the fearful attachment style and
symptom reporting, via the mediating variables of negative
affectivity and alexithymia, is robust across populations.
Although insecurely attached students did not report visiting
their GPs more frequently than did securely attached
students, our findings may have clinical implications when
it comes to interpreting the symptom reports of people who
have visited their doctors.
A limitation of the study was that each attachment style
was measured using only a self-report measure, albeit a
measure which reportedly has good convergent validity with
more complex measures [3], including peer reports, partner
reports, and expert rater measures. The attachment measure
might have been affected by a social desirability bias,
although if that had been the case, we might have expected a
greater number of participants to endorse the dismissive
attachment scale more highly. An interview measure such as
the Family Attachment Interview or the Peer Attachment
Interview (both Bartholomew and Horowitz [10]), although
more time consuming, might have provided a more objective
measure. On the other hand, questionnaires were completed
anonymously, which could be expected to reduce bias due to
social desirability. Lastly, due to the cross-sectional nature of
this study, it is impossible to establish the causality of the
effects observed. However, attachment style has been found
to be relatively stable over many years [35], and therefore, it
287
is likely that attachment style precedes symptom reporting
and can be used as a predictor to some degree.
References
[1] Hazan C, Shaver P. Conceptualising romantic love as an attachment
process. J Pers Soc Psychol 1987;52:511 – 24.
[2] Ainsworth MDS, Blehar MC, Waters E, Wall S. Patterns of attach-
ment: a psychological study of the strange situation. Hillsdale (NJ)7
Erlbaum, 1978.
[3] Griffin DW, Bartholomew K. The metaphysics of measurement:
the case of adult attachment. In: Bartholomew K, Perlman D, editors.
Advances in personal relationships, vol. 5. London7 Kingsley, 1994.
pp. 17 – 52.
[4] Waters E, Merrick S, Treboux D, Crowell J, Albersheim L. Attach-
ment security in infancy and early adulthood: a twenty-year
longitudinal study. Child Dev 2000;3:684 – 9.
[5] Ciechanowski PS, Walker EA, Katon WJ, Russo JE. Attachment
theory: a model for health care utilization and somatisation. Psycho-
som Med 2002;64:660 – 7.
[6] Feeney JA, Ryan SM. Attachment style and affect regulation:
relationships with health behaviour and family experiences of illness
in a student sample. Health Psychol 1994;13:334 – 45.
[7] Kotler T, Buzwell S, Romeo Y, Bowland J. Avoidant attachment as a
risk factor for mental health. Br J Med Psychol 1994;67:237 – 45.
[8] Schmidt S, Strauss B, Braehler E. Subjective physical complaints and
hypochondriacal features from an attachment theoretical perspective.
Psychol Psychother 2002;75:313 – 32.
[9] Wearden A, Cook L, Vaughan-Jones J. Adult attachment, alexithymia,
symptom reporting and health-related coping styles. J Psychosom Res
2003;55:341 – 7.
[10] Bartholomew K, Horowitz LM. Attachment styles among young
adults: a test of a four-category model. J Pers Soc Psychol 1991;
61:226 – 44.
[11] Collins NL, Read SJ. Adult attachment, working models, and
relationship quality in dating couples. J Pers Soc Psychol 1990;
58:644 – 63.
[12] Watson D, Pennebaker JW. Health complaints, stress, and distress:
exploring the central role of negative affectivity. Psychol Rev 1989;
96:234 – 54.
[13] DeGucht V, Heiser W. Alexithymia and somatisation. A quantitative
review of the literature. J Psychosom Res 2003;54:425 – 34.
[14] Taylor GJ, Bagby RM, Parker JDA, editors. Disorders of affect
regulation: alexithymia in medical and psychiatric illness. Cambridge7
Cambridge Univ Press, 1997.
[15] Bagby RM, Taylor GJ. Affect dysregulation and alexithymia. In:
Taylor GJ, Bagby RM, Parker JDA, editors. Disorders of affect
regulation: alexithymia in medical and psychiatric illness. Cambridge7
Cambridge Univ Press, 1997. pp. 26 – 45.
[16] Gergely G, Watson JS. The social biofeedback theory of parental
affect-mirroring: the development of emotional self-awareness and
self-control in infancy. Int J Psychoanal 1996;77:1 – 31.
[17] Bagby RM, Parker JDA, Taylor GJ. The twenty-item Toronto
Alexithymia Scale: I Item selection and cross-validation of the factor
structure. J Psychosom Res 1994;38:23 – 32.
[18] Bagby RM, Taylor GJ. Measurement and validation of the alexithymia
construct. In: Taylor GJ, Bagby RM, Parker JDA, editors. Disorders of
affect regulation: alexithymia in medical and psychiatric illness.
Cambridge7 Cambridge Univ Press, 1997. pp. 46 – 66.
[19] Taylor GJ, Bagby RM, Parker JD. The alexithymia construct A
potential paradigm for psychosomatic medicine. Psychosomatics
1991;32:153 – 64.
[20] Kooiman CG. The status of alexithymia as a risk factor in medi-
cally unexplained physical symptoms. Compr Psychiatry 1998;
39:152 – 9.
288 A.J. Wearden et al. / Journal of Psychosomatic Research 58 (2005) 279–288
[21] De Gucht V, Heiser W. Alexithymia and somatization A quantitative
review of the literature. J Psychosom Res 2003;54:425 – 34.
[22] Troisi A, D’Argenio A, Peracchio F, Petti P. Insecure attachment and
alexithymia in young men with mood symptoms. J Nerv Ment Dis
2001;189:311 – 6.
[23] Rosenberg M. Society and the adolescent self image. Princeton (NJ)7
Princeton Univ Press, 1965.
[24] Rief W, Heuser J, Fichter MM. What does the Toronto Alexithymia
Scale TAS-R measure? J Clin Psychol 1996;52:423 – 9.
[25] Deary IJ, Scott S, Wilson JA. Neuroticism, alexithymia and medically
unexplained symptoms. Pers Individ Differ 1997;22:551 – 64.
[26] Kooiman CG, Spinhoven P, Trijsburg RW. The assessment of
alexithymia: a critical review of the literature and a psychometric
study of the TAS-20. J Psychosom Res 2002;53:1083 – 90.
[27] Watson D, Clark LA, Tellegen A. Development and validation of brief
measures of positive and negative affect: the PANAS Scales. J Pers
Soc Psychol 1988;54:1063 – 70.
[28] Pennebaker JW. The psychology of physical symptoms. New York7
Springer, 1982.
[29] Robbins JM, Kirmayer LJ. Attributions of common somatic symp-
toms. Psychol Med 1991;21:1029 – 45.
[30] Feeney JA. Implications of attachment style for patterns of health and
illness. Child Care Health Dev 2000;26:277 – 88.
[31] Berembaum H, James T. Correlates and retrospectively reported
antecedents of alexithymia. Psychosom Med 1994;56:353 – 9.
[32] Nichols K, Gergely G, Fonagy P. Experimental protocols for
investigating relationships among mother–infant interaction, affect
regulation, physiological markers of stress responsiveness, and
attachment. Bull Menninger Clin 2001;65:371 – 9.
[33] Carnelly K, Pietromonaco PR, Jaffe K. Depression, working models
of others and relationship functioning. J Pers Soc Psychol 1994;66:
127 – 40.
[34] Sayar K, Kirmayer LJ, Taillefer SS. Predictors of somatic symptoms
in depressive disorder. Gen Hosp Psychiatry 2003;25:108 – 14.
[35] Klohnen EC, John OP. Working models of attachment: a theory-
based prototype approach. In: Simpson JA, Rholes WS, editors.
Attachment theory and close relationships. New York7 Guilford,
1998. pp. 115 – 40.