J. Behav. Ther. & Exp. Psychiat.

44 (2013) 194e200

Contents lists available at SciVerse ScienceDirect

Journal of Behavior Therapy and

Experimental Psychiatry
journal homepage: www.elsevier.com/locate/jbtep

Experimental analysis of the relationship between depressed mood and

compulsive buying
Michael Kyrios a, Paul McQueen b, Richard Moulding a, *
a
Swinburne University of Technology, Melbourne, Australia
b
University of Melbourne, Melbourne, Australia

a r t i c l e i n f o a b s t r a c t

Article history: Background and objectives: Compulsive buying is a serious but understudied problem, where individuals
Received 25 January 2012 are unable to resist or control their buying behaviour, leading to substantial social and financial prob-
Received in revised form lems. To date there has been a lack of experimental research into the disorder.
19 October 2012
Methods: The relationship between mood and compulsive buying was examined in compulsive buyers
Accepted 24 October 2012
(N ¼ 18) and non-clinical controls (N ¼ 17), using experimental information-processing paradigms. In
study 1, it was expected that, if buying behaviours function as a coping strategy for depressed mood, then
Keywords:
an induction of depressed mood would lead to an enhanced memory for appealing consumer-objects in
Compulsive buying
Oniomania
compulsive buyers, but not controls. In study 2, we examined the association between emotional
Impulse control disorders and functional constructs and consumer items. It was expected that compulsive buyers would show
Obsessive compulsive spectrum disorders stronger semantic relationships and thus better episodic memory for objecteemotion pairs, relative to
Memory objectefunction pairs, for appealing items.
Results: Unexpectedly, in study 1 the memory-facilitating effect of depressed mood was evident among
control participants and absent among compulsive buyers. In study 2, compulsive buyers showed a lesser
association of undesirable objects with positive emotional concepts than did non-clinical controls, and
compulsive buyers were found to more strongly associate all consumer items with emotional concepts
than with concepts of function.
Limitations: Key limitations were low power and possible floor effects due to error frequency data.
Conclusion: These findings provide insights into the processes underlying CB phenomena, in particular
supporting the role of mood in compulsive buying.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction buyers being female in clinical samples, although a community-

Compulsive buying (CB) is a serious and common psychological
problem associated with marital conflict, bankruptcy, theft,
embezzlement and even suicide attempts (Christenson et al., 1994).
Three core features are generally agreed to characterise CB: (a)
subjectively irresistible urges to buy; (b) loss of control over buying
behaviour; and (c) continued compulsive buying behaviour despite
adverse consequences (Dittmar, 2005a). Estimates of the preva-
lence of CB in the US have ranged between 1.8% and 8.1% of the
general population depending on the cut-off used (Faber &
O’Guinn, 1992). Onset is typically between the ages of 18 and 30
(Black, 1996; Dittmar, 2005a), and with around 90% of compulsive
* Corresponding author. Brain and Psychological Research Centre, Faculty of Life
and Social Science, Swinburne University of Technology, PO Box 218, Hawthorn VIC,
Melbourne, Australia. Tel.: þ61 3 9214 4686.
E-mail address: rmoulding@swin.edu.au (R. Moulding).
based prevalence study has suggested that compulsive buying
may be prevalent in similar rates in males and females (Koran,
Faber, Aboujaoude, Large, & Serpe, 2006).
Kyrios, Frost, and Steketee (2004) have developed a cognitive-
behavioural model of compulsive buying (for a compatible model,
see Kellett & Bolton, 2009). The model emphasizes domains such as
maladaptive attachments to objects, specific cognitions about the
uniqueness of, and responsibility for, desired objects, and self
construals (Frost, Kyrios, McCarthy, & Mathews, 2007). Compulsive
buyers exhibit concerns about the potential loss of opportunities
“to get a bargain” or to purchase an object with particular desired
characteristics and, despite perceptions that purchased objects are
essential, they often remain unused after being purchased (Kyrios
et al., 2004). For example, Kyrios, Moulding, Arnold, and Ciorciari
(submitted for publication) found that beliefs about the utility of
items e that items are uniquely able help the individual improve
their lives e such as “This item is so unique that there is nothing

0005-7916/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jbtep.2012.10.004
 M. Kyrios et al. / J. Behav. Ther. & Exp. Psychiat. 44 (2013) 194e200
else like it in the world”, “If I buy this, people will find me more
attractive”, and “This item will make me feel better” e were
strongly related to buying symptoms in a student and community
sample (N ¼ 240) over-and-above mood and other variables
implicated in social models of buying (e.g., materialistic values).
CB operates in a negative cycle with depression, with buying
episodes operating to relieve negative mood in the short-term, but
in the long term the lack of satisfaction from consumer objects and
negative consequences of buying episodes serve to perpetuate
depression. Kyrios et al. (2004) found moderate correlations
between CB and depression symptoms, consistent with the beliefs
identified in their model that buying will serve to alleviate negative
mood or improve self-worth. They also found that compulsive
buyers believed that objects could compensate for negative
emotions, even when controlling for depressed mood per se. Faber
and Christenson (1996) found approximately 25% of normal buyers
indicated that buying alters their mood, compared to 96% of
compulsive buyers. More directly, Miltenberger et al. (2003) found
that inflated pre-buying self-ratings of depression dropped signif-
icantly during buying episodes but increased again post-purchase.
Therefore, some findings suggest that individuals with CB may
deliberately use buying episodes as a dysfunctional method of
regulating negative emotions. However, it has not yet been shown
that experimentally manipulating mood state has an impact on
buying behaviour and associated information processing. One
difficulty in such experimental research is that it is difficult to
establish a controlled experimental environment in which to
examine the influence of modulating negative affect on CB. An
alternative approach is to establish the relationship between mood
and CB by examining information processing. If the “compensation
hypothesis” is correct, then compulsive buyers should associate
consumer objects with emotional concepts. That is, the expectation
that purchasing consumer items will serve to compensate for
unpleasant emotions should lead to individuals with CB having
a stronger semantic relationship between emotional concepts and
consumer items. Experimentally demonstrating the existence of
such associations among individuals with CB, which significantly
differ from the associations of non-buyers, would be strong
evidence in support of the compensation hypothesis. This
hypothesis was examined in two studies using participants with CB
and non-clinical control participants.
2. Study one
In study one, we aimed to investigate the link between negative
mood and buying through examining the effect of a negative or
neutral mood induction on recognition memory for idiosyncrati-
cally desired vs. non-desired items. Items primed in working
memory can lead to a “pop-out” effect in visual search tasks for
stimuli that match some or all elements of the memory prime (Soto,
Humphries & Heinke, 2006). Therefore, if depressed mood is
associated with CB, it will trigger buying-related mental processes
and activate these in working memory, effectively “priming” the
participant to attend to buying-related stimuli. An induced
depressed mood should lead to direct attention to items perceived
to be able to compensate for depressed mood, but not to items with
no perceived compensatory role. This, in turn, would lead to
superior engagement and, therefore, recognition of these items by
people who meet CB criteria, of the CB-related items (items
subjectively perceived as appealing) presented in the memory task.
In this study, individuals with compulsive buying and non-
clinical (NC) controls were exposed to a depressed or neutral
mood induction, prior to being asked to memorise a sequence of
subjectively attractive and unattractive consumer items. They were
then shown the items along with distractor items, and asked to
195
identify which had been shown previously. For the compulsive
buyers but not the control group, highly-ranked items, which
should be more associated with mood-relieving effects, were ex-
pected to be attended to more readily in the memory task when
these associations are primed by mood state (i.e., in the depressed
mood group). Consequently, it was hypothesised that CB partici-
pants but not non-clinical controls would make fewer errors when
asked to recognise high-ranked items in the depressed mood state
relative to the neutral mood state, after the general effect of mood
was accounted for. These effects were expected for miss errors and
not false positives, due to the effects of facilitation being limited to
the association between aroused emotion and the co-occurrence of
an event (Kazui et al., 2000).
2.1. Method
2.1.1. Participants
Participants undertook both study 1 and 2, and were recruited
via advertisements placed around universities, public libraries and
notice boards, and through responses to media publicity including
newspaper articles and television and radio interviews. Interested
participants completed an initial screening questionnaire to
determine suitability (approximately 200 screening questionnaires
were distributed, N ¼ 85 returned). Additional measures were also
given that are detailed in a larger questionnaire-based study
detailed elsewhere (Kyrios, Moulding, & Zabel, submitted for
publication), before participants were invited to participate in the
experimental studies. Participants received AUD$20 reimburse-
ment of travel costs. To qualify for the CB group, individuals had to
obtain scores on the Compulsive Buying Scale below a threshold
of 1.34, identified by Faber and O’Guinn (1992) as the cut-off for
compulsive buying; and they had to endorse at least one of two
screening questions, “I am a compulsive shopper (or ‘addicted’ to
shopping)” or “Currently experiencing problems due to excessive
buying” (6 participants excluded for meeting only one of two
criteria). Individuals who did not meet either of these criteria were
invited to participate in the control group. In addition, individuals
classified as compulsive buyers were required to be below the
threshold for mania and borderline personality disorder and to be
below the extremely high-range for depression on the measures
described below (overall 17 participants met one or more of these
exclusion criteria). No participants included in the study reported
significant psychiatric history besides prior treatment for depres-
sion or anxiety disorders. Overall, 35 participants were invited to
participate in the CB group and 27 in the NC group. Of these, a final
sample of 35 participants (28 female) participated in the experi-
mental tasks, with eighteen in the control group and 17 in the
compulsive buyer group. One participant in the CB group was
excluded from Study 2 due to a computer malfunction. The groups
did not differ on gender, being single, age, or education (see
Table 1).
2.1.2. Questionnaires
2.1.2.1. Altman Self-Rating Mania Scale (AMS; Altman, Hedeker,
Peterson, & Davis, 1997). A 5-item self-rating scale of the severity
of manic symptoms. The scale has satisfactory reliability and
validity.
2.1.2.2. Borderline Screening Measure (BSM; Hyler et al., 1989).
A 9-item scale measuring borderline personality tendencies. The
PDQ-4 is a widely used reliable and valid measure.
2.1.2.3. Compulsive Buying Scale (CBS; Faber & O’Guinn, 1989).
A well-validated 7-item screening measure for CB that has
demonstrated good reliability and validity (Faber & O’Guinn, 1989,
196 M. Kyrios et al. / J. Behav. Ther. & Exp. Psychiat. 44 (2013) 194e200
Table 1
Demographic characteristics of sample.
Variable Value Control
n %
Gender Female 16
Relationship Single 11
Married/de Facto 4 23.5
Separated/divorced 2 11.8
Mean
Age 37.1
Education 15.6
1992). Two of the seven items pertain to emotional reactions to
shopping, while the remaining five relate to financial aspects of
buying.
2.1.2.4. Compulsive Acquisition Scale (CAS; Frost et al., 1998).
An 18-item scale that measures the extent to which individuals
acquire and feel compelled to acquire possessions. The CAS-Buy
subscale is a broad measure of CB behaviour, which has demon-
strated satisfactory reliability in previous research (Frost, Steketee,
& Williams, 2002; Kyrios et al., 2004).
2.1.2.5. Depression Anxiety Stress Scales (DASS; Lovibond & Lovibond,
1995). A 42-item scale measuring depression, anxiety and stress
over the previous week. Excellent internal consistency, 2-week
test-retest reliability for a clinical sample, as well as high discrim-
inant validity between the three subscales have been reported.
2.1.2.6. Visual-analogue Scales (VAS) of happiness, anxiety and
depression. The validity of the mood induction was examined using
three VAS scales. The scales were anchored at 0, 50 and 100, with
written descriptors (e.g., “Not at all happy”, “Extremely happy”).
2.1.3. Procedure
Participants completed the questions and demographic infor-
mation, then studies 1 and 2, before completing a final computer
task described elsewhere (Kyrios, Moulding, & McQueen, in
preparation). In the first task, participants were shown a series of
pairs of photographs of consumer items, and asked to select the
item in each pair that they would be more likely to buy. In total, 56
items were ranked from each of eight categories (body care items,
accessories, sports equipment, kitchen items, clothes, jewellery,
footware, electronic leisure items; derived from Dittmar, Beattie, &
Friese, 1996). The initial order of presentation of item pairs was
randomised, and an iterative implementation of a merge-sort
algorithm was used to rank order the items by preference based
on participants’ selections from, on average, 225 pairs of items.
Subsequently, participants from each group were randomised to
a neutral or depressed mood induction. They completed the three
VAS scales for mood, and were given written instructions for the
mood-induction procedure. The induction involved listening to
a 10-min piece of music while recalling a memory that made them
feel either sad or neutral (Badcock & Allen, 2003; Kulbartz-Klatt,
Florin, & Pook, 1999). Following the mood induction, participants
again completed the VAS scales. The following musical pieces were
selected e “Russia Under the Mongolian Yoke” by Prokofiev played
at half speed (induction of depressed-mood e ID group); and
“Waltzes numbers. 11 and 12” by Chopin played at normal speed
(induction of neutral-mood e IN group; Badcock & Allen, 2003;
Wood, Saltzberg, & Goldsamt, 1990).
For the memory task, the 56 consumer items from the ranking
task were split into two 28-item sets by ranked pairs. Oddly ranked
Compulsive buying Test
n %
88.9 12 70.6 c2(1) ¼ 0.87, p ¼ 0.35
64.7 7 58.6 c2(1) ¼ 0.01, p ¼ 0.93
4 33.3
1 8.3
(SD) Mean (SD)
13.7 38.9 15.7 F(1,26) ¼ 0.12, p ¼ 0.73
4.1 14.9 3.8 F(1,26) ¼ 0.16, p ¼ 0.69
items were placed in a distractor set, while evenly ranked items
were placed in a target set. Following the mood induction, partic-
ipants were shown the 28-items from the target set in a random
order, with each shown for two seconds, and they were instructed
that they would be asked to subsequently recall which items they
had seen. Following this, participants were immediately presented
with each of the original 56 items, in random order, and asked to
identify the target items. Note that while the memory task was
novel to this experiment, as it is specific to memory for compulsive
buying-related items, it is based on a standard 2-alternative forced-
choice recognition memory task (e.g., Tulving, 1981). Following
this, participants completed the three VAS scales again, and were
debriefed on the mood-induction.
2.2. Results
2.2.1. Statistical methods
All analyses were performed using R (R Development Core
Team, 2005), with the NLME package (Pinheiro, Bates, DebRoy, &
Sarkar, 2005). Most analyses were ANOVAs to test significance of
terms in linear mixed-effects (LME) models built using a hierar-
chical stepwise reduction approach. Linear mixed-effects models
are an extension of the standard linear model, which are robust for
unbalanced data, as was the case in this study (Pinheiro & Bates,
2000). For each analysis, a full-factorial model was first con-
structed including all first-order factors, covariates and interaction
terms adjusted for any violations of assumptions. In order to
maximize power, models were then subjected to stepwise removal
of terms based on the summary table of the marginal significance of
terms (based on Doig et al., 1998), until each term in the model
exhibited significance, or model fit was found to deteriorate based
on when the AIC value increased and ANOVA tests of likelihood
ratios revealed significant changes to the model.
2.2.2. Validity of groups
Supporting the validity of the groups, the compulsive buying
group reported significantly more CB symptoms on the compulsive
buying measures (CBS, M ¼ 2.75, M ¼ 1.23 respectively;
F(1,31) ¼ 70.55, p < 0.0001; CAS-buy, M ¼ 61.12, M ¼ 31.75,
F(1,29) ¼ 43.52, p < 0.0001). The CB group also demonstrated
greater levels of depression, F(1,31) ¼ 7.83, p ¼ 0.009 (M ¼ 11.88,
M ¼ 3.72); anxiety, F(1,31) ¼ 7.48, p ¼ 0.010 (M ¼ 8.88, M ¼ 2.39);
and stress, F(1,31) ¼ 9.17, p ¼ 0.005 (M ¼ 17.24, M ¼ 7.78).
2.2.3. Manipulation check of mood induction
To check the mood induction was effective, a 2 (group e CB vs
control)  2 (mood e induced depression vs neutral)  3 (time e
pre, induced mood; post-experiment) model was constructed.
Overall the groups did not differ, F(2,31) ¼ 0.34, p ¼ 0.563. However,
the mood induction was effective, with main effects of mood e
F(1,31) ¼ 8.50, p ¼ 0.007, qualified by a significant interaction
 M. Kyrios et al. / J. Behav. Ther. & Exp. Psychiat. 44 (2013) 194e200
(i.e., pre-post mood induction), F(4,67) ¼ 3.68, p ¼ 0.009; post-hoc
tests indicating that the pre-time point showed lower depression
than the post-time points, but only for the induced depression
group. While there was a significant group by group  mood
interaction, F(1,31) ¼ 11.69, p ¼ 0.002, the group  mood  time
term did not contribute to the model variance, suggesting that the
CB group demonstrated the same general response to the mood
inductions as the NC group. Equivalent results were found for
happiness, with a significant main effect of mood induction,
F(1,33) ¼ 18.44, p < 0.001, and induction by time interaction,
F(4,66) ¼ 4.98, p ¼ 0.014. Analyses for anxiety did not show any
significant effects, indicating that the manipulation was specific for
depression/happiness.
2.2.4. Experimental findings for errors
Experimental results were analysed by group (control vs. CB),
mood (induced depression vs neutral), object attractiveness rank
(low, middle, high e based on respective thirds of the participant’s
preference for the object), and error type (false positive, or miss).
After reduction, the final model found no differences for group,
F(1,31) ¼ 2.01, p ¼ 0.166, group  mood, F(2,31) ¼ 1.97, p ¼ 0.157, or
mood  error type, F(2,171) ¼ 0.27, p ¼ 0.763; item-ranking was not
included after reduction of the model. There was a significant third-
order interaction of group  mood  error type, F(2,171) ¼ 5.35,
p ¼ 0.006. Therefore, fewer miss errors were made by control but
not experimental participants in the induced depression vs. neutral
group. Planned post-hoc analyses indicated that NCs made signif-
icantly fewer miss errors in the ID group than in the IN group,
t(31) ¼ 2.43, p ¼ 0.021. No significant difference emerged in the rate
of miss errors for CBs, and their trend was to make more miss errors
in the ID group than in the IN group, t(31) ¼ 1.08, p ¼ 0.288 (see
Fig. 1). This contradicted the hypothesis, which predicted that there
would be a third-order interaction effect of group  mood  rank,
with CBs making fewer miss errors for highly-ranked items when
depressed than when in a neutral mood.
2.3. Discussion of study one
The compensation hypothesis for CB suggested there should be
a significantly lower proportion of miss errors than false positives
in the CB group, especially in the depressed mood condition for
appealing items. The differences in performance on the memory
task between the NC and CB groups that emerged were indepen-
dent of the appeal of items but dependent on the mood condition.
Fig. 1. Interaction plots of significant interaction effects accounting for errors in study one.
197
When the neutral mood groups of NCs and CBs were compared, the
performance was largely indistinguishable. Unexpectedly, what
distinguished the groups was that CBs in the depressed condition
made significantly more miss errors than NCs in the depressed
condition.
A possible explanation for this result is that CBs may have low
tolerance for depression and automatically engage in compensa-
tory processes. As discussed, a memory, rather than shopping, task
was chosen to avoid interference of the task content on the induced
mood state. It is possible, however, that viewing consumer images
during the memory task “mimicked” a shopping trip enough to
compensate for CBs’ depressed mood. Anecdotal reports from many
of the compulsive buyers were that the task was “a bit like going
shopping”, and was an element of the study that they tended to
enjoy.
We speculate that CB’s attention in the depressed but not the
neutral condition was directed to mood compensation processes,
distracting from the memory task. This explanation could account
for the paradoxical, though non-significant, increase in miss errors
in the CB group. In terms of the proposed model, this could be taken
to uphold the hypothesis that depressed mood is a trigger for
buying urges, as induced depressed mood was the sole factor that
revealed differences in memory performance between the groups.
For depressed CBs, attention may have been directed to mood-
compensation rather than task performance, leading to poorer
memory for presented items relative to the control group. Future
research is needed to investigate this possibility. In addition, it
should be noted that the CB group generally reported worsened
mood relative to the NC group. While the mood-induction was
found to affect both groups similarly, future studies with larger
sample sizes should control for overall differences in mood
between group as differences in general mood may affect the
results found (rather than differences being due to CB per se).
3. Study two
Study two aimed to extend the examination of the relationship
between mood and objects using a semantic-memory task. Dittmar
(2005a) found that psychologically motivated purchases were
associated with an expectation that the object would improve self-
image in about 17% of cases, and that in approximately 48% of cases
positive mood change was experienced immediately following the
purchase. This powerful association between objects, self-image
and mood should produce semantic networks which over time
198 M. Kyrios et al. / J. Behav. Ther. & Exp. Psychiat. 44 (2013) 194e200
directly link certain consumer objects with emotional concepts.
Given that prior semantic relationships between pairs of items can
facilitate the episodic memory of those items being paired together
(Silberman, Miikkulainen, & Bentin, 2001), it was hypothesised that
the semantic relationship between an object and an emotion could
be demonstrated by measuring individuals’ memories for pairs of
images of objects along with emotion words, compared to pairs of
object images with function words.
Participants were shown consumer items paired with words
from four categories (positive emotional, negative emotional,
functional and neutral), and had to subsequently identify these
pairings in a recognition memory task. As all groups were expected
to possess semantic associations between objects and functions,
but not between objects and unrelated words, fewer errors were
predicted for objects paired with functional words rather than with
neutral words across all participants and groups. As control
participants do not engage in psychologically-motivated buying as
regularly as compulsive buyers, the semantic relationship between
emotions and objects was expected to be very distal (weak) for
controls, regardless of the subjective appeal (rank) of the item,
whereas associations with the utility, or function, of objects should
be more proximal. For the control group it was therefore expected
that episodic memory for objectefunction pairs would be better
than the memory for objecteemotion pairs, regardless of item rank.
In contrast, CBs were expected to exhibit better memory for
objecteemotion pairs than objectefunction pairs for highly
appealing items. For non-appealing items, which are not typically
purchased during compulsive buying episodes, the same pattern as
controls was predicted. Since buying episodes involve both nega-
tive emotions before, and positive mood change during and after
buying (Dittmar, 2005b; Miltenberger et al., 2003), both negative
and positive emotions were expected to be associated with highly-
desired items for compulsive buyers, with positive emotions being
more strongly associated.
3.1. Method
3.1.1. Procedure
Participants were shown two sequences of 12 images of
consumer items, paired with words. The images were of the six
items most highly ranked by the participant and the six lowest-
ranked items, with each item displayed twice as a target, each
time paired with a different target word. Across the two sequences,
six words from each of four word categories (positive emotional
[þE], negative emotional [E], functional [FC] and neutral [NU])
were paired with the items, balanced for rank. Functional words
were words that could generically refer to items as having a func-
tion, being effective, useful or practical. Neutral words were words
that could refer to an object, but with neither emotional nor
functional significance. Word-image pairs were displayed for four
seconds, followed by a one second blank square. Following each 12-
image sequence, participants were shown 24 wordeimage pairs,
half from the previous sequence and half of which included the
same words and images in random pairings. Participants indicated
whether they had seen each pairing.
3.2. Results
3.2.1. Model building
Data was analysed using the same procedure as study 1, based
on the three variables of group (CB or control), word type (þE, E,
FC), and rank (low, high). Inclusion of “neutral” words led to
substantial deterioration of the models and the category was
therefore excluded from analyses. The final models included the
terms group, group  rank and group  rank  category. Of these
effects, group was not significant, F(1,32) ¼ 2.78, p ¼ 0.105; and
neither was group by rank, F(2,160) ¼ 0.06, p ¼ 0.942. However, the
third order interaction, group  rank  category was significant,
F(8,160) ¼ 2.23, p ¼ 0.028 (Fig. 2).
3.2.2. Experimental results
The key difference between the NC and CB groups was the rate
of errors made on low-ranked items paired with positive emotional
words, which were significantly lower in the NC group, t(32) ¼ 2.10,
p ¼ 0.044 (Fig. 2). The lowest rate of errors for CBs was for low
ranked items paired with negative emotional words, but this did
not differ significantly between the NCs and CBs, t(32) ¼ 0.43,
p ¼ 0.668. The CBs made more errors than NCs when items were
paired with functional words, regardless of rank, although this
difference only approached significance, t(64) ¼ 1.98, p ¼ 0.052.
While the patterns of errors differed according to rank in both the
NC and CB groups, a post-hoc ANOVA on the model for only the
control participants indicated no significant differences in error
rates by item rank, F(1,85) ¼ 0.00, p ¼ 1.000, or word categories by
rank, F(4,85) ¼ 0.38, p ¼ 0.821. Thus, functional and emotion-pairs
had equivalent error rates in the NC group, contrary to predictions.
CBs’ rate of errors depended on both item rank and item pairs.
For high-ranked (HR) items, negative or positive emotion pairs
were associated with less errors than functional word pairs,
consistent with the hypothesis, t(160) ¼ 1.63, p ¼ 0.003. For low-
ranked (LR) items, functional and emotional pairs had equivalent
rates of errors, t(160) ¼ 0.22, p ¼ 0.828, as in the NC group. It was
predicted that the fewer errors among CBs would be made on HR
items paired with positive emotional words vs. negative emotional
words, but this was not borne out, t(160) ¼ 0.60, p ¼ 0.545. Positive
word pairs were however associated with less errors in the CB
group for HR vs. LR data, t(160) ¼ 2.83, p ¼ 0.017.
3.3. Discussion of study two
Study two was designed to demonstrate that the regular linking
of consumer objects to both negative and positive emotional states
during psychologically-motivated buying episodes would lead to
the formation of stronger semantic associations for CBs. It was
hypothesised that, among the control participants, errors would be
lower for items paired with functional words than items paired
with emotional words. In practice, for the NCs, the rate of errors
was generally invariant across all the categories of words, sug-
gesting that emotional and functional words have equidistant
semantic associations to consumer items among the general pop-
ulation. This suggests that some emotional impact is given to
consumer objects even in a normal group, consistent with dimen-
sional views of buying that suggest that emotional buying occurs
even in non-clinical individuals, but these associations did not
predominate over associations with the functionality of items.
Examination of the post-hoc tests indicated that, relative to the
“baseline” performance of the NC group (where the category of
words and rankings had no bearing on the rate of errors), emotional
and functional words did show differential associations with
consumer items among the CB group. Compared to NCs, CBs had
a trend towards making more errors for all items paired with
functional words (p ¼ 0.052), made more errors for low-ranked
items paired with positive emotional words, and essentially the
same rate of errors for all items paired with negative emotional
words and highly-ranked items paired with positive emotional
words. Thus, these patterns suggest some differences between
groups according to proximal associations with semantic descrip-
tors. In particular, CBs seem less likely to associate objects with
functional concepts, consistent with descriptions of compulsive
buying as often not emphasizing the normal function of objects, but
 M. Kyrios et al. / J. Behav. Ther. & Exp. Psychiat. 44 (2013) 194e200
Fig. 2. Plots of significant interaction effects in study two.
rather purchasing for mood or identity-repair reasons. Further-
more, CBs made significantly more errors than the NC group in
remembering non-desirable items paired with positive emotional
words, and this rate was also higher than when desirable items
were paired with positive emotional words for the CBs. This
suggests that positive emotions have a semantically more distal
association with non-desirable items than they do for desired
objects for CBs. It was hypothesised that there would be a similar
pattern of association by CBs as NCs for these non-desired items,
assuming that non-desired items do not become linked with
consumer objects through psychologically-motivated buying
episodes. Perhaps, though, items at the lowest end of CBs’ rankings
were not just undesired, but actively rejected. This was reflected by
some CB participants’ verbal comments during the ranking task,
such as, “That is so ugly!”
Where CBs were presented with highly desirable items, pairings
with either positive or negative emotional words were remem-
bered more readily than pairings with functional words. The
strongest effect was for desirable items paired with positive
emotional words, although there was no significant difference
between the rate of errors for high-ranked items paired with
positive and negative words. Overall the results support the notion
that a semantically more proximal relationship exists between
desirable consumer items and both positive and negative
emotional words than functional words for CBs, as predicted.
4. Overall discussion
Together, these findings provide insights into the processes
underlying CB phenomena. The importance of mood in CB is sup-
ported by our findings; although future replication studies would
strengthen the generalizability of our research. The unexpected
finding from study one of greater rates of miss errors rates by
depressed compulsive buyers compared to depressed controls was
taken to indicate that attention was being directed to recruiting the
presented consumer items into the mood-repair process, which led
to a negative impact on memory. The results of study two provide
evidence for the role of depressed mood in producing urges to buy.
Better memory among compulsive buyers, regardless of the appeal
of items, for emotions linked to consumer items, compared to
concepts of function linked to items, can be understood in terms of
the temporal associations between specific items and moods.
Shopping episodes may include “browsing” behaviour, with acute
urges to buy then occurring and leading to purchases. The browsing
process will bring the compulsive buyer into contact with both
199
wanted and unwanted items. Thus, associations will be formed
between mood and consumer items in general. Purchases, accom-
panied by positive mood change, however, only occur for particular
items that, as per the findings of Dittmar et al. (1996), are typically
purchased during compulsive buying episodes.
If replicated, the results of these studies suggest that it is
possible to design effective therapeutic interventions that could
occur within the controlled environment of the therapist’s
consultation room. Study one was taken to suggest that a mood-
repair process was occurring within the laboratory setting, even
though the situation was substantially different to an actual
shopping trip. Study two provides evidence for a particular area in
which cognitive-behavioural interventions may be useful. Since
emotional concepts were found to be strongly associated with
consumer items, treatment involving exposure to consumer items
coupled with both cognitive and behavioural techniques intended
to extinguish maladaptive associations between objects and
emotions could prove effective. Mitchell, Burgard, Faber, Crosby,
and de Zwaan (2006) have demonstrated positive results from
a CBT treatment program that includes thought restructuring and
exposure and response prevention as components of treatment.
Study two suggests that the thought restructuring component of
such therapies should place some emphasis on extinguishing the
associations between consumer items and emotions. The studies
also reinforce the use of techniques such as the development of
strategies for dealing with high-risk times for shopping (e.g.,
alternative pleasurable activities for when emotions are negative;
e.g., see Mitchell, 2011).
The studies presented here were the first to experimentally
manipulate factors such as mood in order to study compulsive
buying. A key limitation of this research is the low number of
participants in each group, with consequentially low power for
statistical analyses. As much as possible, the impact of this limita-
tion was minimised by utilising linear mixed-effects models that
allowed the removal of terms that did not strongly contribute to the
variance in the observed data. Such removal of terms increased the
power of the final models, reducing the probability of type two
errors across the remaining terms. Nevertheless, the possibility
exists when using this approach that removed terms may have
shown significance with a larger sample. In addition to low
participant numbers, the use of frequency data in the studies
produced floor effects. There was sufficient variability in studies
one and two for statistical analysis but the frequency of errors did
not ideally follow a normal distribution. Finally, future studies
should also attempt to control for differences in general mood, and
200 M. Kyrios et al. / J. Behav. Ther. & Exp. Psychiat. 44 (2013) 194e200
perhaps in hoarding behaviour, that may co-exist with differences
in compulsive-buying symptoms.
Overall, the present studies support that future research of CB
could further incorporate experimental methodologies. For instance,
future research could examine differences in cognitive and behav-
ioural patterns between subtypes of buyers, for example between
individuals who are compulsive buyers vs. those who just overspend
(e.g., perhaps due to a lack of budgeting skills). Furthermore, future
research could experimentally manipulate “virtual” shopping or
buying trips to examine the interrelations between mood and
buying urges or related phenomena in more ecologically-valid
contexts. Such methodological advances could help overcome
some of the limitations of questionnaire-based research.
Declaration of interest
All authors declare that they have no conflicts of interest.
Role of funding organizations
The study was not financially supported by a sponsor.
Acknowledgements
The authors would like to thank all the individuals who
participated in this study, and the two anonymous reviewers for
their suggestions.
References
Altman, E. G., Hedeker, D., Peterson, J. L., & Davis, J. M. (1997). Altman self-rating
mania scale. Biological Psychiatry, 42, 948e955.
Badcock, P. B. T., & Allen, N. B. (2003). Adaptive social reasoning in depressed mood
and depressive vulnerability. Cognition and Emotion, 17, 647e670.
Black, D. W. (1996). Compulsive buying: a review. Journal of Clinical Psychiatry, 57,
50e54.
Christenson, G. A., Faber, R. J., de Zwaan, M., Raymond, N. C., Specker, S. M.,
Ekern, M. D., et al. (1994). Compulsive buying: descriptive characteristics and
psychiatric comorbidity. Journal of Clinical Psychiatry, 55, 5e11.
Dittmar, H. (2005a). Compulsive buying e a growing concern? An examination of
gender, age, and endorsement of materialistic values as predictors. British
Journal of Psychology, 96, 467e491.
Dittmar, H. (2005b). A new look at “compulsive buying”: self-discrepancies and
materialistic values as predictors of compulsive buying tendency. Journal of
Social and Clinical Psychology, 24, 832e859.
Dittmar, H., Beattie, J., & Friese, S. (1996). Objects, decision consideration and
self-image in men’s and women’s impulse purchases. Acta Psychologica, 93,
187e206.
Doig, C. J., Sutherland, L. R., Sandham, J. D., Fick, G. H., Verhoef, M., & Meddings, J. B.
(1998). Increased intestinal permeability is associated with the development of
multiple organ dysfunction syndrome in critically ill ICU patients. American
Journal of Respiratory and Critical Care Medicine, 158, 444e451.
Faber, R. J., & Christenson, G. (1996). In the mood to buy: differences in the mood
states experienced by compulsive buyers and other consumers. Psychology &
Marketing, 13, 803e819.
Faber, R. J., & O’Guinn, T. C. (1989). Classifying compulsive consumers: advances
in the development of a diagnostic tool. Advances in Consumer Research, 16,
738e744.
Faber, R. J., & O’Guinn, T. C. (1992). A clinical screener for compulsive buying. Journal
of Consumer Research, 19, 459e469.
Frost, R. O., Kim, H., Morris, C., Bloss, C., Murray-Close, M., & Steketee, G. (1998).
Hoarding, compulsive buying and reasons for saving. Behaviour Research and
Therapy, 36, 657e664.
Frost, R. O., Kyrios, M., McCarthy, K., & Mathews, Y. (2007). Self-ambivalence
and attachment to possessions. Journal of Cognitive Psychotherapy, 21,
236e247.
Frost, R. O., Steketee, G., & Williams, L. F. (2002). Compulsive buying, compulsive
hoarding and obsessive-compulsive disorder. Behaviour Research and Therapy,
33, 201e214.
Hyler, S. E., Rieder, R. O., Williams, J. B., Spitzer, R. L., Hendler, J., & Lyons, M.
(1989). A comparison of clinical and self-report diagnoses of DSM-III
personality disorders in 552 patients. Comprehensive Psychiatry, 30,
170e178.
Kazui, H., Mori, E., Hashimoto, M., Hirono, N., Imamaura, T., Tanimukai, S., et al.
(2000). Impact of emotion on memory: controlled study of the influence of
emotionally charged material on declaritive memory in Alzheimer’s disease.
British Journal of Psychiatry, 177, 343e347.
Kellett, S., & Bolton, J. V. (2009). Compulsive buying: a cognitive-behavioural model.
Clinical Psychology and Psychotherapy, 16, 83e99.
Koran, L. M., Faber, R. J., Aboujaoude, E., Large, M. D., & Serpe, R. T. (2006). Estimated
prevalence of compulsive buying behaviour in the United States. American
Journal of Psychiatry, 163, 1806e1812.
Kulbartz-Klatt, Y. J., Florin, I., & Pook, M. (1999). Bulimia nervosa: mood changes do
have an impact on body width estimation. British Journal of Clinical Psychology,
38, 279e287.
Kyrios, M., Frost, R. O., & Steketee, G. (2004). Cognitions in compulsive buying and
acquisition. Cognitive Therapy & Research, 28, 241e258.
Kyrios, M., Moulding, R., Arnold, F., & Ciorciari, J. Cognitive, self, and mood factors in
compulsive buying, submitted for publication.
Kyrios, M., Moulding, R., & McQueen, P. M. The influence of beliefs on urge to
acquire in compulsive buying: an experimental analysis, in preparation.
Kyrios, M., Moulding, R., & Zabel, P. A comparative study of compulsive buying with
clinical and non-clinical controls, submitted for publication.
Lovibond, P. F., & Lovibond, S. H. (1995). The structure of negative emotional
states: comparison of the Depression Anxiety Stress Scales (DASS) with the
Beck Depression and Anxiety Inventories. Behaviour Research and Therapy, 33,
335e343.
Miltenberger, R. G., Redlin, J., Crosby, R., Stickney, M., Mitchell, J., Wonderlich, S.,
et al. (2003). Direct and retrospective assessment of factors contributing to
compulsive buying. Journal of Behavior Therapy and Experimental Psychiatry,
34, 1e9.
Mitchell, J. E. (2011). Compulsive buying disorder group treatment manual. In
A. Muller, & J. E. Mitchell (Eds.), Compulsive buying: Clinical foundations and
treatment. New York: Routledge.
Mitchell, J. E., Burgard, M., Faber, R. J., Crosby, R. D., & de Zwaan, M. (2006). Cognitive
behavioral therapy for compulsive buying disorder. Behaviour Research and
Therapy, 44, 1859e1865.
Pinheiro, J. C., & Bates, D. M. (2000). Mixed-effects models in S and S-plus. New York:
Springer.
Pinheiro, J. C., Bates, D. M., DebRoy, S., & Sarkar, D. (2005). nlme: Linear and nonlinear
mixed effects models. R package version 3.1-66.
R Development Core Team. (2005). R: A language and environment for statistical
computing (v. 2.2.1). Vienna: R Foundation for Statistical Computing. http://
www.R-project.org.
Silberman, Y., Miikkulainen, R., & Bentin, S. (2001). Semantic effect on episodic
associations. In J. D. Moore, & K. Stenning (Eds.), Proceedings of the twenty-third
annual conference of the Cognitive Science Society (pp. 934e939). Mahwah, NJ:
Erlbaum.
Soto, D., Humphreys, G. W., & Heinke, D. (2006). Working memory can guide pop-
out search. Vision Research, 46, 1010e1018.
Tulving, E. (1981). Similarity relations in recognition. Journal of Verbal Learning and
Verbal Behavior, 20, 479e496.
Wood, J. V., Saltzberg, J. A., & Goldsamt, L. A. (1990). Does affect induce self-focused
attention? Journal of Personality and Social Psychology, 58, 899e908.