Journal of Psychosomatic Research 63 (2007) 663 – 671
Memory bias for health-related information in somatoform disorders
Alexandra Martin4, Anika Buech, Christina Schwenk, Winfried Rief
Section for Clinical Psychology and Psychotherapy, Philipps-University Marburg, Marburg, Germany
Received 1 June 2006; received in revised form 3 April 2007; accepted 1 May 2007
Abstract
Objective: Cognitive processes are considered to be relevant to
the etiology and maintenance of somatoform disorders (SFDs). The
aim of this study was to assess explicit and implicit information-
processing bias for disorder-congruent information in SFDs.
Methods: A clinical sample of 33 patients suffering from multiple
somatoform symptoms (SSI-3/5) and 25 healthy controls per-
formed an encoding task with computer-presented word lists
(illness related, negative, positive, neutral content), subsequently
Keywords: Information-processing bias; Cognitive model; Memory; Somatoform disorder; Somatization
Introduction
Somatoform disorders (SFDs), according to the Diag-
nostic and Statistical Manual of Mental Disorders, Fourth
Edition, Text Revision (DSM-IV-TR) and the International
Classification of Diseases, Tenth Revision (ICD-10), cover
a heterogeneous range of conditions, all of them sharing
the central feature of bodily symptoms that cannot be
fully explained by any medical factor. The impact of
SFDs on the health care system is tremendous, as SFDs
are among the most prevalent psychiatric disorders and,
moreover, are associated with severe impairment in
important areas of functioning and with high health care
utilization [1–4].
In current models of SFDs, information-processing
aspects are considered to be relevant [5–8]. These include
an abnormal amplifying perceptual style [5], restrictive
4 Corresponding author. Section for Clinical Psychology and Psycho-
therapy, Philipps-University, Gutenbergstr. 18, D-35032 Marburg,
Germany. Tel.: +49 6421 282 3656, +49 6421 282 3657; fax: +49 6421
2828904.
E-mail address: martin@staff.uni-marburg.de (A. Martin).
0022-3999/07/$ – see front matter D 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jpsychores.2007.05.005
followed by explicit memory tests (free recall and recognition) and
an implicit test (word-stem completion). Results: The somatoform
group showed a memory bias for illness-related stimuli in the
word-stem completion task, whereas the two groups did not differ
in explicit memory tests. This effect could not be explained by
comorbid depression. Conclusion: These results provide some
support for current theories on SFDs.
D 2007 Elsevier Inc. All rights reserved.
assumptions about health and body functions [9], and an
enduring tendency to misinterpret bodily sensations and
other health-related information as evidence of serious
physical illness [10]. The interaction of perceptual and
interpretative biases of ambiguous body signals can result
in a vicious circle of anxiety, physiological arousal, and
intensification of symptoms (Fig. 1). Furthermore, organic
causal beliefs and vulnerability attributions have been
found to be associated with dysfunctional illness behavior
[11], which in turn might contribute to the process of the
syndrome becoming chronic.
In cognitive–behavioral theories of SFDs, it has been
outlined that beliefs about physical sensations signaling
serious illnesses increase anxiety concerning health, which
can then lead to selective cognitive biases favoring
information that confirms illness belief while discounting
information that contradicts it. Cognitive biases can occur
at different stages of information processing, either at the
encoding-of-information stage, indicated by an attentional
(or preattentional) bias for disorder-relevant stimuli
(bintegrative processQ), or at the point of volitional
retrieval of information from memory, indicating deeper
elaboration of schema-congruent information (belaboration
664 A. Martin et al. / Journal of Psychosomatic Research 63 (2007) 663 – 671
Fig. 1. The cognitive–psychobiological model of SFDs (modified from Rief
and Nanke [6]).
processQ). Aspects of information processing have also
contributed to the understanding of mood and anxiety
disorders. Williams et al. [12,13] provided a cognitive
model of information processing in order to account for
different mnemonic and attentional biases found in these
disorders (e.g., Refs. [14–16]). According to Williams et
al., anxiety is associated with automatic encoding of
fearful stimuli, whereas depression is more likely asso-
ciated with elaboration of negative emotional stimuli.
Accordingly, the bias occurring at integration stage can be
revealed by implicit tests such as perception-based or
word-stem completion tasks, whereas biases related to
elaboration stages of information processing can be
revealed by explicit memory tests (e.g., free recall or
recognition of previously learned material).
The influence of memory in SFDs has been inves-
tigated in a few studies only, with most of them studying
explicit memory effects. In regard to chronic pain
disorders, a couple of findings suggest a memory bias
for pain-related information (sensory words) [17]. How-
ever, these findings should be generalized to SFDs with
caution, as many of the studies investigated chronic pain
conditions with some kind of pathophysiological findings
(e.g., arthritis), and information processes might differ
from SFDs. Results from studies regarding hypochondria-
sis are mixed. Durso et al. [18] could not demonstrate
biases for health-related information with a recognition
task in hypochondriac students. However, in a study by
Brown et al. [19], hypochondriacal individuals did not
show a perceptual bias for health-related information but
did, however, demonstrate evidence suggesting a memory
bias: Within-group comparisons showed that two hypo-
chondriacal samples recalled more health-related words
than nonhealth words, whereas the effects in control
samples were not significant. Pauli and Alpers [20]
reported that patients with hypochondriasis and somato-
form pain disorder and patients with hypochondriasis
without pain disorder showed an enhanced immediate
recall of pain-related words.
Only very few studies assessed memory bias in subjects
suffering from multiple somatoform symptoms, such as in
somatization disorder or undifferentiated SFDs. Rief et al.
[21], using a word category decision task and subsequent free
recall, did not demonstrate a disorder-specific explicit
memory bias in these subjects. Scholz et al. [22] also did
not find an explicit information-processing bias based on an
auditorily presented lexical decision task in students with
multiple somatoform symptoms. However, their results
suggested an implicit information-processing bias for threat-
ening health-related words. In a recent study, a group of SFD
patients showed an explicit memory bias for physical threat
words (the free-recall ratio of physical threat words was
higher than that in controls), as well as supraliminal inter-
ferences for physical threat words in the emotional Stroop
task; the study did not, however, reveal an implicit memory
bias in the tachistoscopic word-identification task [23].
Overall, the number of experimental studies on
information-processing bias in SFDs, especially in soma-
tization disorders or related groups with multiple somato-
form symptoms, is still very limited, and existing studies
have focused mainly on elaborative processes, revealing
inconsistent results. Therefore, the aim of the present
study was to assess memory biases for disorder-congruent
stimuli in SFDs. A clinical sample of patients suffering
from multiple somatoform symptoms and healthy controls
performed an encoding task with computer-presented word
lists (illness related, negative, positive, neutral content),
subsequently followed by standard explicit memory tests
(free recall, recognition) and an implicit test (a word-stem
completion task). We hypothesized that individuals with
multiple somatoform symptoms would show a memory
bias favoring health-related information in explicit and
implicit tasks. As comorbidity with depression is high in
SFDs, we additionally controlled for its influence in
secondary analyses.
Method
Sample
Somatization syndrome group
Subjects were included in the clinical sample only if
multiple somatoform symptoms were present as per the
criterion of the bsomatization indexQ (SSI-3/5), which was
assessed with Screening for Somatoform Symptoms 2
(SOMS-2). According to this criterion, men had to report
at least three somatoform symptoms and women had to
report at least five somatoform symptoms as being present
during the past 2 years (SOMS-2) and which could not be
explained by medical/organic factors. The SSI-3/5 criterion
corresponds to Escobar et al.’s SSI-4/6 proposal to classify
somatization syndrome [24].
Control group
Healthy controls did not fulfill the SSI-3/5 criterion for
somatization syndrome. If any symptoms were reported
 A. Martin et al. / Journal of Psychosomatic Research 63 (2007) 663 – 671
(SOMS-2), they needed to be rated as being medically
explained by their physicians (SOMS-2 Item 55: yes,
reported symptoms could be explained with a medical
explanation by their physician).
The total sample comprised 58 subjects: 33 in the
somatoform group (SFDs) and 25 in the healthy control
group. The majority of subjects of the study’s clinical
sample were patients with SFDs who contacted the Depart-
ment of Clinical Psychology and Psychotherapy and the
outpatient clinic for various reasons (n=23). Participants of
the somatization syndrome group were recruited with an
advertisement in local papers (n=10). The control group
comprised volunteers who were recruited mainly by means
of an advertisement requesting for volunteers for a
psychological experiment.
Assessment/questionnaires
SOMS-2 [24,25] is a self-report instrument for the
detection of SFDs that was used as the measure of
somatization syndrome (yes/no) in the present study. It lists
53 bodily symptoms covering all somatoform symptoms
mentioned in DSM-IV-TR and ICD-10 as occurring in
somatization disorders. Subjects were asked to indicate the
symptoms that were present during the past 2 years and for
which physicians had not been able to find a clear organic
cause. An additional 15 items directly refer to classification
criteria, such as disability due to symptoms, frequent doctor
visits, onset, and duration. Item 55 asks for a confirmation
that the previously indicated symptoms could not be
explained by any known medical condition. The psycho-
metric properties of SOMS-2 are well established, with
good test–retest reliability (72 h; r=. 85), high internal
consistency (.88), high discriminant validity, high sensitivity
(98%), and acceptable specificity (63%), to correctly
identify patients with a clinically relevant somatization
syndrome (as defined by the somatization syndrome SSI-4/6
of Escobar et al. [26]). Correlations between SOMS and
structured clinical interviews varied between .72 and .82 for
various somatization indices.
The state version of Screening for Somatoform Symp-
toms 7 (SOMS-7) was used to identify the number and the
severity of somatoform symptoms that subjects suffered
from during the last 7 days. The scale has shown a high
internal consistency (a=.92) and sensitivity to change
following treatment [27]. The mean severity score varies
between 0 and 4 (only symptoms of at least moderate
disability were counted for the current symptom number).
The Whitely Index (WI) [28,29] is a self-report
questionnaire widely used for assessing degrees of hypo-
chondriac anxiety and concern, with excellent reliability and
validity [30,31]. The 14-item version with dichotomous
response alternatives was used, with the total score varying
between 0 and 14.
The Beck Depression Inventory (BDI) [32,33] was used
as a measure to assess the severity of depressive symptoms.
665
While its psychometric properties have not yet been
established explicitly in SFDs, a range of studies have
shown high internal consistencies (mean a=.87) and
moderate to high concurrent validities (r’sz.6–.76) of the
BDI in psychiatric and nonpsychiatric samples (e.g.,
medical and chronic pain patients) [34].
Intelligence quotient (IQ) was assessed with a vocabu-
lary test (Mehrfachwahl-Wortschatz-Intelligenztest (MWT-
B) [35]. MWT-B is an economic measure of global
crystallized intelligence. The instrument has been shown
to be valid in psychiatric populations and is recommended
for sample descriptions [36]. In the present study, it was
used to control for comparable IQ levels across the clinical
and control groups.
Materials
Four types of word stimuli were used in the memory
tasks: two types of emotionally negative words [illness
related (e.g., breathlessness, infection, stroke) and general
negative words (e.g., failure, jealousy, war)], emotionally
positive words (e.g., affection, confidence, compliment),
and neutral words (e.g., ballpen, carpet, paper). Construc-
tion of the word list material was based on word stimuli
previously used in comparable experiments [20,22,37],
self-report questionnaires [38,39], and expert inputs,
resulting in a primary word list of 200 words, with 50
words in each word category. This primary word list was
presented to 50 adult students of multiple disciplines who
were asked to rate the category (illness related, emotionally
negative, emotionally positive, neutral) of each word. Only
those words for which at least 80% of the subjects rated
the correct category were considered for further inclusion.
Each illness-related word was matched with one word of
each of the other categories with respect to word length
[number of syllables (mean/S.D.): illness-related 3.00/0.67,
negative 3.00/0.67, positive 3.00/0.67, neutral 2.83/0.75]
and word frequency (illness-related 12.8/4.6, negative
12.8/5.4, positive 12.8/4.6, neutral 12.8/4.6; according to
http://wortschatz.uni-leipzig.de).
Based on these results, 48 completely parallelized
words (12 per category) were chosen for the encoding
and free-recall task (Appendix). In addition, 48 partially
parallelized words (12 per category) were selected as
unprimed material for the recognition task. The word-stem
completion task consisted of 44 of 48 words of the encoding
task (11 words per category). Three words were excluded as
their word stems did not offer an alternative word
completion with a comparable word frequency (negative:
ignorance; positive: tenderness; neutral: encyclopedia),
and one word was excluded because the word stem
could be completed by either an illness-related word or a
positive word.
In addition, the independent Student’s sample (n=50)
rated the degree of threat of all illness-related and negative
words on a 4-point scale (0=not at all, 3=very much) to
666 A. Martin et al. / Journal of Psychosomatic Research 63 (2007) 663 – 671
ensure the comparable emotional valence of these two word
types. The mean emotional valence of illness-related words
(mean=2.59, S.D.=0.44) did not differ from the mean
valence of negative words (mean=2.56, S.D.=0.43)
[t(49)=0.57, PN.05].
Procedure
Encoding and free recall
A random selection of three words from each category
was subsequently assigned to four blocks (A, B, C, D) in
random order. One additional filling word was presented at
the beginning and one additional filling word was presented
at the end of each block to avoid primacy and recency
effects; thus, the blocks comprised 14 words each. This
within-block word order was kept stable for the encoding
task. Four-word stimuli blocks were then combined into
four presentation sequences (ABCD, BCDA, CDAB,
DABC) so that each block was presented with the same
frequency at each position. Sequence presentations were
balanced across all subjects, and its possible effect was
analyzed (see Results). The words were presented in white
letters (font type: Arial; font size: 48) against a black
background on a computer screen (15 in.; flat-panel display;
1024768-pixel resolution) positioned 70 cm in front of
seated subjects.
Each task was preceded by written instructions on the
screen, which were rephrased by the investigator. Prior to
each encoding block, the subjects were instructed to
memorize the words by relating them to subjectively
relevant events. The subjects started the presentation of
the words individually. Each word was presented for 7 s.
Each of the four encoding blocks was followed by a 3-min
distraction task (a visual search for differences between two
almost identical drawings). The first explicit memory task
(free recall) took place thereafter. During a 5-min period,
subjects were asked to write down all words they
remembered. Subjects had to complete the sequence of
bencoding–distraction task–free recallQ four times.1
Word-stem completion
The next task consisted of a word-stem completion test
(implicit memory). A list of 44 word stems (three or four
letters of encoding words in random order) was presented,
and subjects were asked to complete the letters by
choosing the first word that came into mind within a
maximum of 5 min.
1
Control for word order effects: Recall rates regarding absolute word
position in the encoding sequence differed significantly [ F(47, 2632)=2.7,
Pb .01]. Graphical control clearly demonstrated higher recall rates for the
first word of each block (one per category), suggesting a primacy effect.
When these four stimuli were eliminated from analysis, the word order
effect disappeared [ F(43, 2408)=1.33, PN.10)].
Recognition
Afterwards, a word list of 48 primed words and 48
unprimed words (24 of each word type in random order)
was presented. Subjects were instructed to indicate the
words they remembered to have been presented during the
encoding phase. Again, a maximum of 5 min was allowed to
accomplish the last explicit memory task.
Questionnaires
In order to avoid bcontaminatingQ the effects of ques-
tionnaire items, subjects filled in the questionnaire battery
(MWT-B, SOMS-2, SOMS-7, WI and BDI) after the
memory tasks.
Statistical analysis
Dependent variables
Free recall : the number of correctly recalled words per
category and block was counted for each
subject; the total number of correctly
recalled words per category was entered
for analysis.
Recognition : the variable dV (sensitivity index; derived
from signal detection analysis) was computed
from the bhit rateQ and bfalse alarm rateQ
variables of each word type for each subject
(correctly identified words minus false-pos-
itive words).
Word-stem completion : the dependent variable was
derived from the number of words completed
correctly according to the priming list per
category.
Analysis
Recall data were subjected to analysis of variance, with
group as the between-subjects factor and word type as the
within-subject factor. Two (Group)four (Word Type)
analyses of variances were conducted. In case of significant
main effects of the word type factor, planned contrasts
(Helmert) were conducted to compare the memory of
illness-related words with those of the other three categories.
In case of a significant interaction (according to our
hypotheses), subsequent multivariate analyses of variance
(MANOVA) for each group were computed to clarify the
nature of the interaction. The significance level of global
statistical tests was Pb.05 for planned contrasts according to
directed hypotheses (one tailed). All MANOVA were based
on Greenhouse–Geisser correction of degrees of freedom.
Effect sizes were computed as partial g 2 values or Cohen’s
d. Additional analyses of covariance (ANCOVA) were
conducted to control for the possibly confounding effect of
group differences in depression.
Post-hoc power analyses for MANOVA ( f=0.25, a=.05,
N=59) revealed a power of .97 for medium between-factor
 A. Martin et al. / Journal of Psychosomatic Research 63 (2007) 663 – 671 667

Table 1
Sociodemographic and clinical characteristics
Characteristics SFD group (n=33) Control group (n=25) Statistics
Sociodemographic
Gender (female/male) 24/9 16/9 v 2(1)=0.51
Age in years [mean (S.D.)] 49.58 (10.54) 45.88 (10.2) t(56)=1.34
Education (b10 years/z10 years) 9/24 1/24 v 2(1)=5.404
Employed (yes/no) 22/11 15/10 v 2(1)=0.27
Intelligence (MWT-B) [mean (S.D.)] 119.52 (14.04) 124.88 (9.49) t(55)= 1.62
Clinical [mean (S.D.)]
SOMS-2, somatoform symptoms, last 2 years 15.00 (6.57) 1.96 (2.11) t(56)=9.52444
SOMS-7, somatoform symptoms, last 7 days 5.36 (4.19) 0.60 (1.47) t(56)=5.43444
SOMS-7, symptom intensity, last 7 days 0.38 (0.23) 0.05 (0.10) t(56)=6.59444
WI, health anxiety, last 7 days 5.15 (3.68) 1.63 (1.76) t(56)=4.41444
BDI, depression severity, last 7 days 17.26 (14.2) 7.35 (9.06) t(55)=3.0444
4 Pb.05.
44 Pb.01.
444 Pb.001.

effects, .89 for within-factor effects, .31 for medium-
interaction effects, and .70 for large-interaction effects
( f=0.40).
Differences in sociodemographic and descriptive clinical
variables were determined using t tests, where appropriate,
and chi-square test for nominal data. All analyses were
conducted using SPSS statistical package (version 12.0;
SPSS Inc., Chicago, IL, USA).
Results
Subject characteristics
Gender ratio and mean age were comparable between
groups (Table 1). Although the two groups differed with
respect to school education, with a higher number of control
subjects having attended school for at least 10 years, there
were no differences in current working status and mean
verbal intelligence level (MWT-B).
As expected, the SFD group showed a higher number
of lifetime and current somatoform symptoms, higher
symptom severity, more health anxiety, and higher
depression scores.
for unspecific negative and positive words. However,
contrary to our hypotheses, the SFD group did not show
better recall of illness-related words compared to
controls; the interaction GroupWord Type was not
significant [ F(3,168)=0.26].
To rule out possible confounding effects of the sequence
of block presentation, we conducted a secondary analysis of
variance with bsequenceQ as additional between-subjects
factor. Results did not show any significant effect of the
presentation sequence [main effect, F(3,50)=1.25; interac-
tions: Word TypeSequence, F(9,150)=0.90; Word Type-
GroupSequence, F(9,150)=1.32; all P’sN.15].
Word-stem completion
The implicit memory test found no main group effect
[ F(1,56)=1.23]. The main effect Word Type was significant
[ F(3,168)=10.05, Pb.001, partial g 2=.15]. Contrast showed
that illness-related word stems were significantly more often
correctly completed than the word stems of the other three
categories [ F(1,56)=10.59, Pb.01, partial g 2=.16].
The interaction GroupWord Type was significant
[ F(3,168)=2.36, Pb.05, partial g 2=.04]. To clarify the

Free recall
Table 2
Results of memory tasks: number of words, by category per group
The results of the memory tasks are shown in Table
2. The first explicit memory task consisted of free recall Word type [mean (S.D.)]
of all words presented during the encoding task (primed Participant group Illness related Negative Positive Neutral
word list). The SFD group and the control group did Free recall
not differ in their overall recall [ F(1,56)=1.51]. There SFD group (n=33) 5.97 (2.08) 4.91 (2.99) 4.42 (2.56) 5.61 (2.61)
was a main effect for Word Type [ F(3,168)=9.06, Control group (n=25) 6.52 (2.40) 5.28 (2.09) 5.12 (2.44) 6.56 (2.82)
Word-stem completion
Pb.001, partial g 2=.14], showing that the overall free SFD group (n=33) 4.15 (1.97) 3.45 (1.64) 2.45 (1.25) 3.12 (1.32)
recall of illness-related words was significantly better Control group (n=25) 3.16 (1.86) 3.04 (1.27) 2.44 (1.15) 3.20 (1.58)
than any of the other three-word categories [ F(1,56)= Recognition dV
13.01, Pb.01, partial g 2=.19]. Paired comparisons SFD group (n=33) 9.97 (1.51) 8.79 (2.13) 7.82 (2.95) 9.21 (2.32)
showed higher recall rates for illness-related words than Control group (n=25) 9.52 (1.69) 8.32 (2.51) 7.52 (2.88) 9.16 (2.14)
668 A. Martin et al. / Journal of Psychosomatic Research 63 (2007) 663 – 671
Fig. 2. Word-stem completion in the SFD group and in the healthy
control group: the number of words (and standard error) completed
per category.
nature of the interaction (Fig. 2), separate MANOVA were
carried out for each group. In the SFD group, the effect
Word Type was highly significant [ F(3,96)=11.67,
Pb.001, partial g 2=.27]. Furthermore, specific contrasts
confirmed that the number of correctly completed words
was higher for illness-related words than for the other
three categories for the SFD group [ F(1,32)=16.16,
Pb.001, partial g 2=.34]. Paired comparisons confirmed
higher completion rates in the illness-related category than
in any of the other categories ( P’sb .05). Control subjects
did not show a bias for completing illness-related word
stems more often than a bias for completing other word
stems [ F(3,72)=2.15].
Additional t tests were conducted to compare the two
groups with respect to each word category and showed that
the SFD group completed more illness-related words
correctly compared to the control group [t(56)= 1.94,
Pb.05, d=0.52]. The groups did not differ with respect to
the completion of negative, positive, and neutral words
[t’s(56)=0.05–1.05].
The relative implicit bias for illness-related words
(correct completion of illness-related words divided by the
overall correct completion100) correlated with clinical
characteristics of SFDs, but did not show any significant
association with the number of lifetime somatoform
symptoms (r=.08), the number of current somatoform
symptoms (r=.05), symptom severity (r=.04), and the
degree of health anxiety (r=.06).
Post-hoc analysis for unprimed word-stem completions
To rule out a general response bias for illness-related
information in SFD subjects, we conducted the following
post-hoc analysis. An independent researcher (blinded to
study aims and design) categorized all nonhit words in the
word-stem completion task, so that the number of unprimed
words for each category could be determined. The numbers
[mean (S.D.)] of illness-related/negative/positive/neutral
words were 2.2 (1.4)/3.6 (1.7)/5.6 (2.0)/17.3 (3.9) in the
SFD group and 2.3 (1.4)/4.4 (1.8)/4.9 (2.2)/18.0 (3.7) in the
control group. Results of MANOVA with unprimed words
did not show a higher rate of illness-related words in SFD
subjects than in controls; the interaction GroupWord Type
was not significant [ F(3,168)=1.1, P=.36]. This indicates
that a general response bias cannot account for the results of
the word-stem completion task.
Recognition
The second explicit memory task consisted of the
recognition of previously presented words among a word
list of primed and unprimed words. The SFD group and the
control group did not differ in their overall ability to
correctly identify the material (dV) [ F(1,56)=0.40]. How-
ever, once again, a significant main effect of Word Type was
detected [ F(3,168)=19.46, Pb.001, partial g 2=.26]. The dV
for the recognition of illness-related words was significantly
better than the dV for the other word categories [contrast
F(1,56)=33.19, Pb.001, partial g 2=.37]. The overall recog-
nition rate for illness-related words was higher than that for
words with unspecific negative, positive, or neutral content
(paired comparisons, P’sb.05).
Contrary to the hypotheses, it could not be shown that
SFD group subjects recognized illness-related words more
often than did control group subjects [GroupWord Type,
F(3,168)=0.23].
Reanalysis to control for depression severity
In a second step, a reanalysis controlling for BDI
depression (ANCOVA) was conducted. The effects of the
covariate were not significant in the free recall, recog-
nition, and word-stem completion tasks [ F’s(1, 54)=0.01–
0.53]. The pattern of results across the three memory
tasks was the same as that reported in previous analyses:
no group main effects: F’s(1, 54)=0.49–1.2; significant
Word Type main effects: F(3,162)=3.4–12.2 for all tasks;
interaction GroupWord Type being significant in the
word-stem completion task only: F(3,162)=2.30 ( Pb.05,
partial g 2=.04.) Thus, the identified effects of the main
analyses cannot be solely explained by differences in
depression severity.
Association between memory tasks
Free-recall and recognition rates as assessed with the two
explicit memory tasks correlated significantly with each
other (r’s=.42–.67; illness related, r=.42; negative, r=.54;
positive, r=.67; neutral, r=.61). As expected, correlations
between the implicit task and the explicit tasks bfree recallQ
(r’s=.16–.31; illness related, r=.16; negative, r=.24; pos-
 A. Martin et al. / Journal of Psychosomatic Research 63 (2007) 663 – 671
itive, r=.31; neutral, r=.25) and brecognitionQ (r’s=.20–.41;
illness related, r=.34; negative, r=.20; positive, r=.41;
neutral, r=.34) were lower.
Discussion
The main findings of the study were that individuals with
multiple somatoform symptoms showed a bias for illness-
related information in the implicit memory task, whereas
such a bias could not be revealed with the free-recall and
recognition tasks, which require more elaborative processes.
The overall memory capacity (the number of words
correctly remembered) did not differ between the two
groups in any of the tests.
In the word-stem completion task, the subjects completed
word stems with the first word that came into their mind.
Individuals with somatization syndrome filled in the
previously presented illness-related content more often than
controls. This implies that SFD subjects’ first reaction
towards an ambiguous stimulus showed the tendency to
primarily consider body-related and illness-related informa-
tion. In this respect, the finding gives support to cognitive–
behavioral models of SFDs [6,40,41] and hypochondriasis
[42,43]. All of these models emphasize the interaction of
attentional focus on health-related information and body
signals with misinterpretation of these stimuli in a cata-
strophizing manner. The role of memory is not explicitly
formulated, but can be embedded in these models [44]:
cognitive schemata about body functions, health, and illness
can influence the memory, perception, and interpretation of
associated stimuli.
The implicit bias for illness-related information was
found on a categorical level, while its association with
somatoform symptoms severity did not appear to be a linear
one (as indicated by very low correlations).
However, other explanations for the findings in the
word-stem completion task need to be considered. Stem
completion task performance might not reflect pure
implicit information processing but also explicit processes
because participants are sometimes aware that the words
they are reporting were presented during the encoding
task [45]. Based on the methods used in this study, we
were unable to disentangle the relative contribution of
implicit and explicit processes to the observed memory
task performance. Future studies should therefore
consider methods such as the bmethod of oppositionQ
(e.g., see Richardson-Klavehn and Gardiner [46]) to
separate the relative contributions of implicit and
explicit processes.
The results of the implicit memory task might also be
influenced by a general response bias in SFD patients due
to a higher familiarity with illness-related stimuli. In the
present study, the word stems were derived from the
primed word list only, and we did not assess a baseline
669
for word completion prior to the encoding–free recall–
word-stem completion sequence, which was kept constant
across all subjects. We therefore conducted additional
post-hoc analysis with unprimed word-stem completions,
and results did not indicate that SFD subjects had a
general tendency to complete word stems with disorder-
congruent information.
Until now, there have only been a few experimental
studies on memory bias in SFDs, and overall results have
been inconsistent.
Our result of an implicit task-based bias for health-related
information confirms one previous finding. Scholz et al.
[22] reported an implicit, but not explicit, information-
processing bias in subjects with multiple somatoform
symptoms. Contrary to our result, Lim and Kim [23], using
a tachistoscopic word-identification task, did not show
differential patterns of implicit memory across the SFD,
depressive, and panic groups. The absence of implicit
memory biases in this study may be explained by the nature
of this task. As Watkins [47] has argued, an implicit mood-
congruent memory bias does not appear to exist when
perceptually driven tests are used.
We were unable to demonstrate a bias for illness-related
words with the explicit memory tasks of free recall and
recognition. This is in accordance with some previous
findings [18,21,22] but is in contrast to some others
[19,20,23]. Rief et al. [21] did not show a recall bias for
illness-related words in a sample of patients suffering from
multiple somatoform symptoms. The characteristics of the
clinical sample were similar to those of the somatization
syndrome group in our study. Two other studies that
provided some support for an explicit bias included
samples characterized by hypochondriac anxiety [19,20].
It is possible that the inconsistent findings are a result of
sample characteristics, as the group of SFDs covers a
range of heterogeneous disorders; in other words, perhaps
the variety of SFDs suggests different and varied
information-processing biases. Current cognitive models,
however, do not sufficiently differentiate between diag-
noses within the spectrum of SFDs, as has been shown in
anxiety disorders [48].
The previously mentioned study of Lim and Kim [23]
showed a higher free-recall ratio of physical threat words
for SFD patients to physical threat words for healthy
controls. However, memory results were confounded by
reduced memory capacity in SFD patients compared to
controls. In our study, both samples showed comparable
memory capacity, and the overall recall rate appeared to
be higher than that in Lim and Kim. Thus, these
variations might account for the different results on
explicit memory bias. Further studies are needed to
appraise the role of explicit memory processes across
variations in stimulus sets and experimental paradigms in
various SFDs.
670 A. Martin et al. / Journal of Psychosomatic Research 63 (2007) 663 – 671
In the present study, we chose an established
paradigm to assess disorder-specific information-process-
ing bias. However, one might object that confrontation
with word stimuli, followed by subsequent memory
tasks, lacks some ecological validity. It is possible that
memory biases in SFDs appear especially with current
ambiguous events. For example, a recent study showed
that patients with somatization syndrome remembered
increased likelihood estimates of medical causes for body
symptoms, which had been presented in audiotaped
medical records [49].
Further shortcomings of the present study need to be
mentioned. The clinical sample consisted of a group of
patients suffering from somatization syndrome (SSI-3/5),
and the diagnostic procedure relied on self-report data.
Although the instrument (SOMS-2) has been shown to be
considerably sensitive to the detection of SFDs [24,25], a
diagnostic procedure using an interview and medical
assessment to rule out organic symptom causation would
have been preferable. Against the rather liberal criterion
of intake we used in our study, the mean number of
somatoform symptoms (lifetime symptoms, 15; current
symptoms, 5) was much higher. The diagnostic group
characterized by multiple somatoform symptoms is gen-
erally considered to be a very relevant group in the health
care system [2,50].
Even though the sample size of our study was
comparable to or even higher than those in most reference
studies [8–20,22], it limits statistical power to detecting
small effects. However, inspection of the recall and
recognition rates per group and category does not at all
suggest any interaction between group and word type.
Interpretation of the results is further limited by the
absence of a clinical control group. Therefore, it cannot be
determined whether the obtained results were specific to
SFDs. This needs to be investigated in future studies. To
rule out a confounding influence of depression, we
reanalyzed the data; the result pattern persisted even when
controlling for depression.
The choice of the stimulus material for the present study
was subject to a rigorous process: word lists were matched
for word length and frequency, and emotional valence was
comparable for illness-related and unspecific negative
words. Because of ground effects in previous studies, we
asked subjects to memorize the words by relating them to
subjectively relevant events. This encoding strategy could
have been more effective in SFD patients as they
presumably had more past health-related events. Still,
both groups showed comparable memory capacity, and
especially the two explicit tasks did not reveal a
memory bias in the SFD group. In our study, memory
for illness-related words was contrasted with memory
for general negative, positive, and neutral words. To
investigate the specificity of a bias for (threatening)
illness-related stimuli, it might be worth including
positive health-related stimuli in future studies.
To summarize, the results suggest a disorder-congruent
information-processing bias in somatization syndrome,
appearing in a task that reflects more automatic encoding
and does not need deep elaboration. A general tendency to
better recall disorder-related information in explicit mem-
ory tests was not shown. Overall, the results provide some
support for the cognitive model of SFDs. The identifica-
tion of particular maladaptive information-processing
strategies could lead to a better understanding of the
nature of SFDs. However, as the body of evidence is still
very limited and inconclusive, there is still an obvious
need for further research.
References
[1] Barsky A, Ettner SL, Horsky J, Bates DW. Resource utilization of
patients with hypochondriacal health anxiety and somatization. Med
Care 2001;39:705 – 15.
[2] Creed F, Barsky A. A systematic review of the epidemiology of
somatisation disorder and hypochondriasis. J Psychosom Res
2004;56:391 – 408.
[3] Jyv7sj7rvi S, Joukamaa M, V7is7nen E, Larivaara P, Kivel7 S-L,
Kein7nen-Kiukaanniemi S. Somatizing frequent attenders in primary
health care. J Psychosom Res 2001;50:185 – 92.
[4] Nanke A, Rief W. Zur Inanspruchnahme medizinischer Leistungen bei
Patienten mit somatoformen Stfrungen. Psychotherapeut 2003;
48:329 – 35.
[5] Barsky AJ. Amplification, somatization, and the somatoform disor-
ders. Psychosomatics 1992;33:28 – 34.
[6] Rief W, Nanke A. Somatization disorder from a cognitive–psycho-
biological perspective. Curr Opin Psychiatry 1999;12:733 – 8.
[7] Warwick HMC. A cognitive–behavioural approach to hypochondria-
sis and health anxiety. J Psychosom Res 1989;33:705 – 10.
[8] Salkovskis PM, Warwick HMC. Meaning, misinterpretations, and
medicine: a cognitive–behavioural approach to understanding health
anxiety and hypochondriasis. In: Starcevic V, Lipsitt DR, editors.
Modern perspectives on an ancient malady. Oxford7 University Press,
2002. pp. 203 – 22.
[9] Barsky AJ, Coeytaux RR, Sarnie MK, Cleary PD. Hypochondriacal
patient’s beliefs about good health. Am J Psychiatry 1993;150:1085 – 9.
[10] Rief W, Hiller W, Margraf J. Cognitive aspects in hypochondriasis and
the somatization syndrome. J Abnorm Psychol 1998;107:587 – 95.
[11] Rief W, Nanke A, Emmerich J, Bender A, Zech T. Causal illness
attributions in somatoform disorders—associations with comorbidity
and illness behavior. J Psychosom Res 2004;57:367 – 71.
[12] Williams JMG, Watts FN, MacLeod C, Mathews A. Cognitive
psychology and emotional disorders. New York7 Wiley, 1988.
[13] Williams JMG, Watts FN, MacLeod C, Mathews A. Cognitive
psychology and emotional disorders. 2nd ed. Chichester7 Wiley, 1997.
[14] MacLeod C, McLaughlin K. Implicit and explicit memory bias in
anxiety: a conceptual replication. Behav Res Ther 1995;33:1 – 14.
[15] Mathews A, Mogg K, May J, Eysenck M. Implicit and explicit
memory bias in anxiety. J Abnorm Psychol 1989;98:236 – 40.
[16] Watkins PC, Mathews A, Williamson DA, Fuller RD. Mood-
congruent memory in depression: emotional priming or elaboration?
J Abnorm Psychol 1992;101:581 – 6.
[17] Pincus T, Morley S. Cognitive-processing bias in chronic pain: a
review and integration. Psychol Bull 2001;127:599 – 617.
[18] Durso FT, Reardon R, Shore WJ, Delys SM. Memory processes and
hypochondriacal tendencies. J Nerv Ment Dis 1991;179:279 – 83.
[19] Brown HD, Kosslyn SM, Delamater B, Fama J, Barsky AJ. Perceptual
and memory biases for health-related information in hypochondriacal
individuals. J Psychosom Res 1999;47:67 – 78.
 A. Martin et al. / Journal of Psychosomatic Research 63 (2007) 663 – 671 671

[20] Pauli P, Alpers GW. Memory bias in patients with hypochondriasis
and somatoform pain disorder. J Psychosom Res 2002;52:45 – 53.
[21] Rief W, Shaw R, Fichter MM. Elevated levels of psychophysiological
arousal and cortisol in patients with somatization syndrome. Psycho-
som Med 1998;60:198 – 203.
[22] Scholz OB, Ott R, Mqller-Sinik K. Beziehungen zwischen
Parametern der impliziten und expliziten Informationsverarbeitung
und psychosomatischen Selbstberichtsmagen. Verhaltenstherapie
1997;7:217 – 25.
[23] Lim S-L, Kim J-H. Cognitive processing of emotional information in
depression, panic, and somatoform disorder. J Abnorm Psychol
2005;114:50 – 61.
[24] Rief W, Heuser J, Mayrhuber E, Stelzer I, Hiller W, Fichter MM. The
classification of multiple somatoform symptoms. J Nerv Ment Dis
1996;184:680 – 7.
[25] Rief W, Hiller W, Heuser J. SOMS—Das Screening fqr Somatoforme
Stfrungen Manual zum Fragebogen [SOMS—The Screening for
Somatoform Symptoms]. Bern7 Huber-Verlag, 1997.
[26] Escobar JI, Rubio-Stipec M, Canino G, Karno M. Somatic Symptoms
Index (SSI): a new and abridged somatization construct—prevalence
and epidemiological correlates in two large community samples.
J Nerv Ment Dis 1989;177:140 – 6.
[27] Rief W, Hiller W. A new approach to the assessment of the treatment
effects of somatoform disorders. Psychosomatics 2003;44:492 – 8.
[28] Pilowsky I. Dimensions of hypochondriasis. Br J Psychiatry 1967;
113:89 – 93.
[29] Rief W, Hiller W, Geissner E, Fichter MM. Hypochondrie: Erfassung
und erste klinische Ergebnisse. Z klin Psychol 1994;23:34 – 42.
[30] Hiller W, Rief W. Internationale Skalen fqr Hypochondrie. Bern7 Hans
Huber, 2004.
[31] Speckens AEM, Van A, Hemert M, Spinhoven P, Bolk JH. The
diagnostic and prognostic significance of the Whitely Index, the
Illness Attitudes Scales and the Somatosensory Amplification Scale.
Psychol Med 1996;26:1085 – 90.
[32] Beck AT, Steer RA. Beck Depression Inventory Manual. San Antonio7
The Psychological Cooperation, 1987.
[33] Hautzinger M, Bailer M, Worall H, Keller F. Beck Depressionsin-
ventar: 2 Aufl. Bern7 Huber, 1995.
[34] Beck AT, Steer RA, Garbin MG. Psychometric properties of the Beck
Depression Inventory: twenty-five years of evaluation. Clin Psychol
Rev 1988;8:77 – 100.
[35] Lehrl S. Mehrfachwahl-Wortschatz-Intelligenztest (MWT-B). Baling-
en7 Spitta Verlag, 1977.
[36] Satzger W, Fessmann H, Engel RR. Liefern HAWIE-R, WST und
MWT-B vergleichbare IQ-Werte? Z Differ Diagn Psychol 2002;
23:159 – 70.
[37] Shaw R. Psychologische und psychophysiologische Aspekte somato-
former Stfrungen. Marburg7 Tectum Verlag, 1996.
[38] Ehlers A, Margraf J. Fragebogen zu kfrperbezogenen Ängsten,
Kognitionen und Vermeidung. Weinheim7 Beltz Test, 1993.
[39] Hathaway SR, McKinley JR, Engel RR. Minnesota Multiphasic Per-
sonality Inventory—Revised (MMPI-2) Manual. Bern7 Huber, 2000.
[40] Kirmayer LJ, Taillefer S. Somatoform disorders. In: Turner SM,
Hersen M, editors. Adult psychopathology and diagnosis. New York7
John Wiley, 1997. pp. 338 – 83.
[41] Sharpe M, Bass C. Pathophysiological mechanisms in somatization.
Int Rev Psychiatry 1992;4:81 – 97.
[42] Barsky AJ, Wyshak GL. Hypochondriasis and somatosensory
amplification. Br J Psychiatry 1990;157:404 – 9.
[43] Salkovskis PM, Warwick HMC. Morbid preoccupations, health
anxiety and reassurance: a cognitive–behavioural approach to
hypochondriasis. J Psychosom Res 1986;24:597 – 602.
[44] Brown RJ. Psychological mechanisms of medically unexplained
symptoms: an integrative conceptual model. Psychol Bull 2004;
130:793 – 812.
[45] Schacter DL, Bowers J, Booker J. Intention, awareness, and implicit
memory: the retrieval intentionality criterion. In: Lewandowsky S,
Dunn JC, Kirsner K, editors. Implicit memory: theoretical issues.
Hillsdale (NJ)7 Erlbaum, 1989. pp. 47 – 65.
[46] Richardson-Klavehn A, Gardiner JM. Retrieval volition and memorial
awareness in stem completion: an empirical analysis. Psychol Res
1995;87:252 – 71.
[47] Watkins PC. Implicit memory bias in depression. Cogn Emot
2002;16:381 – 402.
[48] Coles ME, Heimberg RG. Memory biases in the anxiety disorders:
current status. Clin Psychol Rev 2002;22:587 – 627.
[49] Rief W, Heitmqller AM, Reisberg K, Rqddel H. Why reassurance fails
in patients with unexplained symptoms—an experimental investiga-
tion of remembered probabilities. PLoS Med 2006;3:e269 [7 pp].
[50] Escobar JI, Golding JM, Hough RL, Karno M, Burnam MA, Wells
KB. Somatization in the community: relationship to disability and use
of services. Am J Public Health 1987;77:837 – 40.

Appendix. Word stimuli, by category

Word type [English translation (original German material)]

Illness related Negative Positive Neutral
Illness (Krankheit) Loss (Verlust) Peace (Frieden) Paper (Papier)
Surgery (Operation) Disaster (Katastrophe) Justice (Gerechtigkeit) Entrance (Eintritt)
Disease (Erkrankung) Failure (Versagen) Confidence (Zuversicht) Second (Sekunde)
Drug (Medikament) Insult (Beleidigung) Leisure time (Feierabend) Front door (Haustqr)
Headache (Kopfschmerzen) Ignorance (Ignoranz) Compliment (Kompliment) Encyclopedia (Lexikon)
Stroke (Schlaganfall) Jealousy (Eifersucht) Affection (Zuneigung) Borough (Stadtbezirk)
Infection (Ansteckung) Bitterness (Bitterkeit) Brightness (Helligkeit) Cassette (Kassette)
Migraine (Migr7ne) Ozone hole (Ozonloch) Summertime (Sommerzeit) Carpet (Teppich)
Breathlessness (Atemnot) Pollution (Verschmutzung) Tenderness (Z7rtlichkeit) Ballpen (Kugelschreiber)
Heart attack (Herzanfall) Rage (Wutausbruch) Heartiness (Herzlichkeit) Rooftile (Dachziegel)
Cancer (Krebs) War (Krieg) Beach (Strand) Cupboard (Schrank)
Joint pain Spitefulness Amusement Mashed potatoes
(Gelenkschmerzen) (Geh7ssigkeit) (Vergnüglichkeit) (Kartoffelbrei)