Appetite 74 (2014) 4854

Contents lists available at ScienceDirect

Appetite
journal homepage: www.elsevier.com/locate/appet

Research report

Chocolate cake. Guilt or celebration? Associations with healthy eating

attitudes, perceived behavioural control, intentions and weight-loss
Roeline G. Kuijer , Jessica A. Boyce
Department of Psychology, University of Canterbury, Private Bag 4800, Christchurch 8140, New Zealand

a r t i c l e

i n f o

Article history:
Received 16 July 2013
Received in revised form 5 November 2013
Accepted 8 November 2013
Available online 23 November 2013
Keywords:
Guilt
Ambivalence
Healthy eating
Weight-loss
Chocolate

a b s t r a c t
Food and eating are often associated with ambivalent feelings: pleasure and enjoyment, but also worry
and guilt. Guilt has the potential to motivate behaviour change, but may also lead to feelings of helplessness and loss of control. This study rstly examined whether a default association of either guilt or celebration with a prototypical forbidden food item (chocolate cake) was related to differences in attitudes,
perceived behavioural control, and intentions in relation to healthy eating, and secondly whether the
default association was related to weight change over an 18 month period (and short term weight-loss
in a subsample of participants with a weight-loss goal). This study did not nd any evidence for adaptive
or motivational properties of guilt. Participants associating chocolate cake with guilt did not report more
positive attitudes or stronger intentions to eat healthy than did those associating chocolate cake with
celebration. Instead, they reported lower levels of perceived behavioural control over eating and were
less successful at maintaining their weight over an 18 month period. Participants with a weight-loss goal
who associated chocolate cake with guilt were less successful at losing weight over a 3 month period
compared to those associating chocolate cake with celebration.
2014 Published by Elsevier Ltd.

Introduction
In modern society, food and eating are associated with ambivalent feelings: pleasure and enjoyment, but also worry and concern
(e.g., about weight gain, appearance and health effects) (Rozin,
Bauer, & Catanese, 2003; Rozin, Fischler, Imada, Sarubin, &
Wrzesniewski, 1999; Rozin, Kurzer, & Cohen, 2002). Rozin et al.
(2003) suggest that these worries and concerns may reduce peoples quality of life and be unproductive in terms of health and
weight control. It is striking that, for example, the prevalence of
obesity is much higher in the United States than in France (OECD
Health Data, 2011), while the Americans tend to associate food
more with health, tend to worry more about food, and focus less
on the enjoyment and experience of food than do the French (Rozin
et al., 1999; see also Rozin et al., 2002). The prototypical example
of a food item that elicits ambivalent feelings in many people is
chocolate. The present study examines whether a default association of chocolate cake with either guilt or celebration is associated with healthy or unhealthy eating behaviours, attitudes,
perceived behavioural control and intentions towards healthy eating, and weight change over an 18 month period. In addition, the
study examines whether either default association is productive
or unproductive when trying to lose weight.
Corresponding author.
E-mail address: roeline.kuijer@canterbury.ac.nz (R.G. Kuijer).
0195-6663/$ - see front matter 2014 Published by Elsevier Ltd.
http://dx.doi.org/10.1016/j.appet.2013.11.013

One of the methods used by Rozin et al. (1999, 2003) to assess

ambivalent feelings towards food is to present forced-choice items
to participants and measure what they consider default ways of
thinking about food. Participants are presented with a number of
food items (e.g., fried egg) and are asked to circle the word they
most readily associate the food item with (cholesterol or breakfast). Responses to forced-choice items capture the persons most
salient thoughts about a particular issue. In a sample of students it
was found that, for example, 22% associated chocolate cake more
with guilt than celebration, and 45% associated fried egg more
with cholesterol than breakfast(Rozin et al., 2003). The present
study focuses on the default association with one of these food
items, that is, chocolate cake.
Chocolate is one of the most craved foodstuffs (Rogers & Smit,
2000) and is the prototypical example of a food item that elicits
ambivalent feelings. It is loved for its taste, scent and texture and
is usually regarded as a special treat or a reward, hence the association with celebration. However, its high fat, sugar and energy content is often viewed negatively and the experience of guilt after
consumption is common (e.g., Dewberry & Ussher, 1994; Macht
& Dettmer, 2006; Rodgers, Stritzke, Bui, Franko, & Chabrol, 2011;
Rogers & Smit, 2000). As a result it is often referred to as a forbidden food. From a self-control perspective, eating chocolate can be
seen as a delayed-cost dilemma: positive short-term consequences
(it tastes good now), but negative long-term consequences (it is
fattening) (Baumeister & Heatherton, 1996; Giner-Sorolla, 2001).

R.G. Kuijer, J.A. Boyce / Appetite 74 (2014) 4854

Delayed-cost dilemmas are often associated with positive hedonic

affect (e.g. fun, relaxing) and negative self-conscious affect (e.g.,
guilt, regret) (Giner-Sorolla, 2001).
Theoretically, self-conscious emotions such as guilt can have
adaptive as well as maladaptive consequences. Guilt requires
self-awareness and self-reection in relation to some personal
standard (Tangney, Stuewig, & Mashek, 2007) and is experienced
as immediate feedback in relation to actual behaviour (e.g., overindulging in chocolate cake when on a diet) or anticipated behaviour
(e.g., thinking about eating chocolate cake). Feeling guilty is
unpleasant, hence guilt may function as a punishment cue, motivating corrective action (Tangney et al., 2007). Or in other words,
feelings of guilt make us do the right thing. Public health campaigns aimed at instilling healthy eating habits are often based
on the premise that a guilty conscience will be a motivator to
change behaviour. Feelings of guilt may also aid self-control
because they remind people of their long-term goal and shift their
attention away from immediate temptation (Baumeister &
Heatherton, 1996). Feeling guilty when thinking about or seeing
chocolate cake may help the person realise that giving in to temptation interferes with a desired long-term goal (e.g., losing or maintaining weight, eating a healthy diet). However, if the behaviour
that elicited the guilt cannot be inhibited, or if corrective action
is not (immediately) possible, then feeling guilty may lead to
maladaptive effects such as self-criticism, loss of control, poor
self-esteem (Tangney et al., 2007). For example, guilt after an initial (minor) violation of ones diet may result in loss of control because the dieter believes there is no point in further restraining
their eating as the damage has already been done and corrective
action is useless (what the hell effect; Herman & Polivy, 1984).
Feelings of guilt about eating and food are very common in
eating disordered populations (i.e., people who suffer from anorexia, bulimia or other eating disorders; e.g., Sassaroli et al., 2005).
And although a number of studies have looked at the occurrence
of eating or food-related guilt in non-disordered populations
(e.g., De Witt Huberts, Evers, & de Ridder, 2013; Dewberry &
Ussher, 1994; Rozin et al., 1999, 2003), only a handful of studies
have looked at the correlates of guilt. With two exceptions (Conradt et al., 2008; Giner-Sorolla, 2001) these studies have all looked
at guilt in relation to chocolate consumption, and most point to potential maladaptive consequences. Chocolate-related guilt has
been associated with self-reported dysfunctional eating patterns
(e.g., restrained eating, emotional eating, bulimia), high anxiety,
depression, low self-esteem, neuroticism, body dissatisfaction
and drive for thinness (Cartwright & Stritzke, 2008; Cramer & Hartleib, 2001; Mller, Dettmer, & Macht, 2008; Rodgers et al., 2011).
An experimental study by Macht and Dettmer (2006) showed that
participants who experienced guilt after eating a chocolate bar also
reported less intense positive affect at the same time. Only two
studies found adaptive effects in the food/weight-loss literature.
In an experimental study, Giner-Sorolla (2001; Study 3) showed
that female dieters who were unobtrusively primed with negative
self-conscious emotion words (including guilt) inhibited their food
intake on a subsequent taste test. Conradt et al. (2008) found that
weight-related guilt (guilt concerning eating habits, exercising and
weight-control) measured at baseline was a signicant predictor of
increased problem-focused engagement strategies 6 months later
in a sample of obese individuals.

The present study

Previous research looking at correlates of food or eating-related guilt has almost exclusively focused on dysfunctional eating
patterns and has predominantly used student samples (e.g.,

49

Cartwright & Stritzke, 2008; Cramer & Hartleib, 2001; Rodgers

et al., 2011). The rst aim of this study was to extend previous
research by examining whether chocolate-related guilt is associated with attitudes, intentions and perceived behavioural control
in relation to healthy eating (rather than dysfunctional eating) in
an adult, community sample. To the best of our knowledge this is
the rst study to examine this question and therefore these
hypotheses are non-directional. As outlined above, guilt may
have adaptive as well as maladaptive consequences (Tangney
et al., 2007). If guilt is adaptive and motivates behaviour change,
then participants who associate chocolate cake with guilt should
report more positive attitudes toward healthy eating and stronger intentions to eat a healthy diet in the future compared to
those who associate chocolate cake with celebration. In contrast,
if guilt is maladaptive and related to loss of control and feelings
of helplessness, then participants who associate chocolate cake
with guilt should report lower levels of perceived behavioural
control (i.e., self-efcacy) over healthy eating. Associations with
current eating behaviours will also be examined. Participants
who associate chocolate cake with guilt are expected to report
unhealthier eating behaviours regardless of the potential adaptive or maladaptive consequences of guilt. After all, if ones eating is not unhealthy then there is no reason to feel guilty in
the rst place.
The second aim of the study was to examine whether a default
association of guilt was prospectively related to weight change
over a period of 18 months. To our knowledge, only Conradt
et al. (2008) have examined actual weight change in relation to
guilt and found that higher levels of guilt were unrelated to weight
change over a 6 month period in an obese sample. If guilt motivates behaviour change then participants who associate chocolate
cake with guilt should be better able to maintain their weight over
time. In contrast, if guilt indeed has maladaptive consequences as
Rozin and others suggest (Rozin et al., 1999, 2003) then participants who associate chocolate cake with guilt may be less able
to maintain their weight over time.
The nal aim of the study was to examine the correlates of
chocolate-related guilt in more detail in a subsample of participants with a current weight-loss goal. Although chocolate is a food
that both dieters and non-dieters feel guilty about (King, Herman,
& Polivy, 1987), dieters generally report higher levels of guilt than
do non-dieters (e.g., Cartwright & Stritzke, 2008). This is not surprising as chocolate will generally present a stronger self-control
dilemma to dieters than to non-dieters because of their wish to
lose or maintain weight. We therefore expected that those with a
weight-loss goal would be more likely to report chocolate-related
guilt compared to those without an active weight-loss goal. However, an important question is whether either default association is
productive or unproductive when trying to lose weight. We therefore examined whether associating chocolate cake with guilt or
celebration was related to the amount of desired weight change,
goal importance, and short-term weight-loss success in participants with an active weight-loss goal. Based on previous research
showing that chocolate-related guilt is associated with higher levels of body dissatisfaction and drive for thinness (Rodgers et al.,
2011), we expected participants who associate chocolate cake with
guilt (vs. celebration) to have a more ambitious weight-loss goal,
and rate their goal as more important to them. The literature on
dieting suggests that people with a weight-loss goal may be more
prone to the negative consequences of guilt as they may be more
inclined to feel that corrective action after a transgression is not
possible (cf. Herman & Polivy, 1984). Thus, in a subset of people
with a weight-loss goal, associating chocolate cake with guilt
may be related to less weight-loss success than does associating
chocolate cake with celebration.

50

R.G. Kuijer, J.A. Boyce / Appetite 74 (2014) 4854

Methods
Participants and procedure
Total sample
Originally, participants were recruited to participate in a crosssectional study on health and well-being. Recruitment took place
over a period of 6 months through random delivery of letters about
the study to 3500 homes in Canterbury, New Zealand. All participants gave their informed consent, and the study was approved
by the university human ethics committee. In order to avoid statistical dependency in the data, data from only one person per household was used in the present study (N = 314). Data from 20
participants could not be used because they had a missing value
on the question asking whether they associated chocolate cake
with guilt or celebration. Most of these participants wrote they
never ate chocolate cake or disliked chocolate cake. The nal sample at baseline therefore consisted of 294 participants.
Eighteen months after the rst participants had entered the
study, participants who at baseline expressed willingness to take
part in future research (N = 252) were re-contacted and received
a follow-up questionnaire with 202 participants returning this
questionnaire. Twelve participants did not report their follow-up
weight, bringing the follow-up sample size down to 190. As participants were initially recruited over a period of 6 months, the
follow-up questionnaire was completed 1218 months post baseline (M = 17.11 months, SD = 2.37, range = 1222 months). However, as the mean time since baseline was so close to 18 months
we refer to this follow-up as the 18 month follow-up throughout
this paper.
At baseline, participants were aged between 18 and 86 years
(M = 48.1, SD = 17.9). The sample was predominantly female
(72%) and of European descent (91%). Compared to New Zealand
census gures for this region, people with higher education levels
were somewhat overrepresented (11% no formal school qualication, 31% secondary school qualication, 22% tertiary school qualication, 35% university degree). Mean Body Mass Index (BMI)
(based on self-reported height and weight) was 25.0 (SD = 4.4;
range: 16.847.2), with men (M = 26.0) having a slightly higher
BMI compared to women (M = 24.6, p < .05). Mean BMI in the present sample was slightly below the national average for both men
and women (27.3 and 27.1, respectively, Ministry of Health,
2008). Participants who did not want to participate in future
research (N = 42) or did not return the follow-up questionnaire
(N = 50) (N = 92 in total; 31% of the original sample of 294) were
younger (p < .01), reported unhealthier eating behaviours
(p < .05), had less positive attitudes toward healthy eating
(p < .01), and were more likely to associate chocolate cake with
guilt (p < .05) at baseline, compared to those who did complete
the 18-month follow-up questionnaire. No other signicant differences were found.

Weight-loss sample
Participants who reported wanting to lose weight in the baseline questionnaire, received an additional questionnaire 3 months
after study entry. As we did not want the Christmas holidays to
interfere with the follow-up assessment, only participants recruited before mid-August received the 3 month follow-up questionnaire. Of the 247 participants recruited before mid August,
53% (N = 131) had a weight-loss goal. Of those people, 109 indicated willingness to participate in future research and received
the 3 month follow-up questionnaire. Eight-three participants
returned the questionnaire. Two participants did not report their
follow-up weight, bringing the sample size down to 81. Participants with a weight-loss goal who did not want to participate in

future research (N = 22) or did not return the follow-up questionnaire (N = 26) (N = 48 in total; 37% of the original weight-loss sample of 131) were younger compared to those who completed the
3 month follow-up (p < .01). No other demographic differences or
differences on baseline measures were found.
Measures
For all multi-item scales, items were summed and then averaged. Means and standard deviations of the key variables, as well
as correlations between these variables and demographic variables
are presented in Table 1.
Total sample
Default association: Following Rozin and colleagues (Rozin et al.,
1999, 2003) participants were asked to indicate of which word
they think rst (forced-choice format) when they read the words
chocolate cake: guilt or celebration.
Eating behaviours: Participants were asked to recall their eating
behaviour over the past 2 weeks with 5 items (based on Baker, Little, & Brownell, 2003). Sample items are: In the past two weeks,
on how many days did you eat in a balanced way with a lot of fruit
and vegetables and In the past two weeks, on how many days did
you eat junk food (potato chips, desserts, sweets, candy bars, etc.).
Data from a small validation study showed that the retrospective
recall of the ve eating behaviours correlated highly with a 2 week
diary report of those behaviours (see Kuijer & Boyce, 2012). All
items were scored on a ve point scale ranging from 1 every
day to 5 less than once a week. Items were scored in such a
way that a higher score on the summed scale indicates healthier
eating behaviours (Cronbachs alpha = .67).
Eating evaluation: Following the eating behaviour questions,
participants were asked to give an overall evaluation of how
healthy their eating had been over the previous 2 weeks (1 = not
very healthy, 7 = very healthy) (Baker et al., 2003).
Theory of Planned Behaviour (TPB) variables: Attitude, Perceived
behavioural control and Intention were all assessed with standard
items (Ajzen, 1991; Conner, Norman, & Bell, 2002). Attitude toward
healthy eating was measured with ve bipolar items on a 7-point
scale: For me, healthy eating is . . . good-bad, important-unimportant, boring-interesting, pleasant-unpleasant, useful-useless
(Cronbachs alpha = .78). All items apart from boring-interesting
were reverse scored. Three questions were asked to measure Perceived behavioural control (Cronbachs alpha = .61): participants
were asked How difcult or easy would the following things be
for you? followed by the options: eat in a balanced way with a
lot of fruit and vegetables, eating moderate amounts of food
and stopping when I am full and staying away from junk food
(1 = very difcult, 5 = very easy). Intention was measured with two
items (Cronbachs alpha = .92): In the next four weeks, do you
plan to eat a healthy diet (balanced diet, moderate amounts and
avoiding too much junk food)? and In the next four weeks, do
you expect to eat healthily? (1 = certainly not, 7 = certainly yes).
18 Month weight change: At baseline and at follow-up
(18 months after baseline) participants were asked how much they
currently weighed. A difference score was computed to assess
weight change. A positive score means weight gain and a negative
score weight loss.
Weight-loss sample
Weight-loss status: In the baseline questionnaire, participants
were asked which of the following statements applied to them:
I am trying to lose some weight, I am happy with my current
weight and I am trying to gain some weight. If participants endorsed the rst option they were asked to complete a number of
questions on weight-loss (see below). Participants trying to lose

51

R.G. Kuijer, J.A. Boyce / Appetite 74 (2014) 4854

Table 1
Correlations between key variables in the study and demographic variables.
Sex

Age

Educ.

BMI

Default

Total sample (N = 294)

Eating behaviours
Eating evaluation
Attitudes
PBC
Intention
FU weight D (kg) (N = 190)

.01
.02
.17*
.00
.06
.01

.47***
.42***
.22***
.42***
.35***
.20**

.25***
.20**
.20**
.14*
.22***
.13

.13*
.19**
.16*
.23***
.03
.03

.14*
.21***
.08
.21***
.12*
.17**

Weight-loss sample (N = 131)

Total WL goal (kg)
Importance goal
3 mo WL goal (kg)
3 mo condence
3 mo weight D (kg) (N = 81)
3 mo effort (N = 81)

.03
.03
.06
.07
.09
.03

.19*
.22*
.13
.29**
.18
.02

.14
.01
.15
.08
.10
.22*

.65***
.01
.39***
.04
.04
.04

.06
.24**
.11
.08
.28*
.13

WS
.23***
.29***
.15*
.33***
.21**
.05

(SD)

4.21
5.45
6.03
3.97
6.13
0.84

(0.67)
(1.29)
(0.86)
(0.80)
(1.04)
(4.50)

7.00
6.94
4.26
5.59
0.11
4.06

(5.07)
(2.09)
(2.61)
(2.50)
(2.80)
(2.35)

Note: Sex: 1 = male, 2 = female. Default = default association: 1 = celebration, 2 = guilt. WS = weight-loss status: 1 = no weight-loss goal, 2 = weight-loss goal.
p < .05.
**
p < .01.
***
p < .001.
*

weight were classied as having a current weight-loss goal. The

remaining participants were categorised as not having a current
weight-loss goal. How to best measure weight-loss or dieting status is a controversial issue in the literature (e.g., Witt, Katterman, &
Lowe, 2013), however, most researchers tend to use a multi-item
dietary restraint scale. Although we did not ask about restrained
eating in the baseline questionnaire, we did include an abbreviated
(4 items; Cronbachs alpha = .88) version of the Dietary Intent Scale
(DIS; Stice, 1998) in the 18 month follow-up. Participants with a
weight-loss goal at baseline scored nearly a standard deviation
higher (M = 2.76) on the DIS compared to participants without a
weight-loss goal (M = 2.01), t(199) = 6.13, p < .001, providing validation for our single, dichotomous item to distinguish between
those with and without a weight-loss goal.
Weight-loss goal, importance and condence: Participants who
reported wanting to lose some weight were asked how much
weight they wanted to lose (weight-loss goal) and how important
it was for them to reach their weight-loss goal (1 = not at all important, 10 = very important). Participants were then asked to think
3 months ahead and report how much weight they wanted to have
lost by then (3 month goal) and how condent they were about
reaching their 3 month goal (1 = not at all condent, 10 = extremely
condent).
Three month weight change and effort: Three months after
baseline participants with a weight-loss goal were asked how
much they currently weighed. A difference score was then computed between baseline and 3 month weight to assess weight
change. They were also asked how hard they had tried (effort) to
reach their weight-loss goal over the past 3 months (1 = not at
all, 10 = extremely).
Analyses
Following descriptive and correlational analyses, univariate
analyses of variance (including demographic variables as covariates if they correlated signicantly with the dependent variables)
were conducted to examine differences between participants associating chocolate cake with guilt/celebration in the total sample
and in the weight-loss sample. The assumptions of homogeneity
of group variances and homogeneity of regression slopes for continuous covariates (age, BMI, education) were met in all analyses.
However, the dichotomous covariates (weight-loss status and
sex) were both related to the default association (see Descriptive
analyses), thereby violating the assumption that covariate and

independent variable should be unrelated. In such situations the

covariate should be included as a factor in the analyses with a full
factorial design to examine potential interactions between covariate and independent variable. If no interactions occur, the main
effect for the independent variable (i.e., default association) can
be reliably interpreted. Given the modest sample size, dichotomous covariates were assessed one at the time where appropriate.
Finally, correlations (and partial correlations controlling for covariates) were examined to explore whether any of the baseline
variables were associated with weight change at follow-up (3
and 18 months).
As participants were not originally recruited for a multi-measurement study, there was considerable drop-out between baseline and follow-ups. We therefore analysed the follow-up data
twice, once including only participants who completed both baseline and follow-up (complete case-analysis), and once replacing
missing values using multiple imputation. A disadvantage of complete-case analysis, apart from a reduction in power, is that non-response bias may inuence the results. To address these issues
multiple imputation was used. With this method, missing values
are predicted using existing values on other variables. This process
is performed multiple times and the analyses are then performed
separately on the imputed data sets. Only the pooled parameter
estimates are reported from the multiple runs (Schafer & Graham,
2002; Swaim, Wayman, & Chen, 2004). In the current study, imputed values were estimated based on all available baseline measures. Ten imputed data sets were created.
As outlined previously, the weight-change variables were
change scores. Change scores have been criticised because they
tend to be more unreliable than their component variables. All
analyses examining predictors of weight-change were therefore
also run using hierarchical regression analyses in which Time 2
weight was rst regressed onto Time 1 weight in Step 1 of the analyses, followed by the predictor variable in Step 2. As results with
change scores are easier to interpret than regression analyses
and the analyses showed the same results, the analyses with
change scores are presented in the text.
Results
Descriptive analyses
Twenty-seven percent of the total sample associated chocolate
cake with guilt and 73% with celebration. In line with Rozin et al.

52

R.G. Kuijer, J.A. Boyce / Appetite 74 (2014) 4854

(2003), women were more likely to associate chocolate cake with

guilt (30%) than were men (17%), v2 (1, N = 294) = 5.22, p < .05.
No signicant differences were found with respect to age,
t(292) = 1.76, ns, level of education, t(292) = 1.55, ns, or BMI,
t(284) = 1.39, ns.
About half of the total sample (56%) reported wanting to lose
weight (40% were happy with their current weight and 4% were
trying to gain weight). Women were more likely to have a
weight-loss goal (63%) than were men (37%), v2 (1, N = 294) =
16.99, p < .001. Participants with a weight-loss goal had higher
BMIs (26.7 vs. 22.8), t(284) = 8.21, p < .001, and were younger
(45.5 vs. 51.3), t(292) = 5.78, p < .01, compared to participants
without a weight-loss goal. There were no signicant differences
in level of education, t(292) = 1.55, ns.
There was a signicant relationship between wanting to lose
weight and associating chocolate cake with guilt: of the participants with a weight-loss goal, 37% associated chocolate with guilt
and 63% with celebration. Of the participants without a weightloss goal, only 13% associated chocolate cake with guilt and 87%
with celebration, v2 (1, N = 294) = 21.64, p < .001.
Means, standard deviations and zero-order correlations between demographics and the main variables in the study are presented in Table 1. Age, education, and BMI (but not sex), were
consistently related to most of the eating related variables in the
total sample. Age and BMI were also related to some of the
weight-related variables in the weight-loss sample. In the analyses
of variance described below, demographic variables were included
as covariates if they were signicantly related to the dependent
variable under examination, and sex was included as a factor in
the analysis with attitudes as the dependent variable. Table 1 further shows that in the total sample having a weight-loss goal was
related to unhealthier eating behaviours and evaluation, less positive attitudes, lower levels of perceived behavioural control and
less healthy eating intentions. Weight-loss status was therefore
added as a factor in the analyses of variance with these variables
as dependent variables.

Table 2
Differences between participants associating chocolate cake with guilt and
celebration.
Guilt

Celebration
M

pg2

SE

Total sample
Eating behaviours1,2,3,4
Eating evaluation1,2,3,4
Attitudes1,2,3,4,5
PBC1,2,3,4
Intention1,2,4
18 mo weight D (kg)1 (CC)
18 mo weight D (kg)1 (MI)

SE

4.15
5.20
6.04
3.80
5.97
2.36
2.51

(.076)
(.149)
(.116)
(.092)
(.128)
(.651)
(.651)

4.23
5.54
6.10
4.03
6.16
0.36
0.61

(.037)
(.074)
(.058)
(.046)
(.064)
(.362)
(.464)

0.82
4.34*
0.20
4.86*
1.87
7.19**
6.08*

.003
.015
.001
.017
.006
.037
.035

Weight-loss sample
Total WL goal (kg)1,3
Importance goal1
3 mo WL goal (kg)3
3 mo condence1
3 mo weight D (kg)1 (CC)
3 mo weight D (kg)1 (MI)
3 mo effort 2 (CC)
3 mo effort 2 (MI)

6.95
7.56
4.83
5.76
1.25
1.45
4.43
4.65

(.528)
(.295)
(.354)
(.353)
(.530)
(.764)
(.436)
(.422)

7.04
6.60
3.92
5.49
0.43
0.27
3.88
3.81

(.396)
(.218)
(.268)
(.264)
(.365)
(.393)
(.308)
(.281)

0.02
6.71*
4.14*
0.37
6.81*
4.83*
1.06
2.40

.000
.050
.033
.003
.079
.081
.013
.030

Where appropriate, means were adjusted for covariates (1age, 2education, 3BMI,
Weight-loss status (WS), 5Sex). WS and Sex were included as factors in the analyses: no signicant interaction effects were found between WS and default, or Sex
and default, all Fs < 2.28, ps > .13.
CC = complete-case analysis, MI = multiple imputation.
Total sample: N = 294; 78 guilt/216 celebration, except for CC analysis: N = 190; 45
guilt/145 celebration. Weight-loss sample: N = 131; 47 guilt/84 celebration, except
for CC analysis: N = 81, 26 guilt/55 celebration.
*
p < .05.
**
p < .01.

p < .001.
4

perceived behavioural control and follow-up weight-change remained signicant (r = .15, p < .05) once age, education, and BMI
were controlled for.

Guilt/celebration and weight-loss in the total sample

Guilt/celebration and weight-loss in the weight-loss sample

Table 1 shows that associating chocolate cake with guilt was

related to unhealthier eating behaviours and evaluation, lower
levels of perceived behavioural control, and less healthy eating
intentions at baseline. In a series of 2 (default association)  2
(weight-loss status) analyses of variance (including covariates
where appropriate as outlined previously), two signicant main effects were found for default association (see Table 2): participants
who associated chocolate cake with guilt evaluated their eating as
unhealthier, and were less condent that they could eat healthy in
the future compared to participants who associated chocolate cake
with celebration. There were no signicant main effects for
weight-loss status or signicant interaction effects, all Fs < 2.28,
ps > .13.
Eighteen months after baseline, participants had gained on
average 0.84 kg (SD = 4.50; range: 10 kg weight-loss19 kg
weight-gain). Associating chocolate cake with guilt was related
to an increase in weight (Table 1). Interestingly, having a weightloss goal at baseline was unrelated to weight change. An ANCOVA
(controlling for age) revealed a main effect for default association.
This effect was signicant regardless of whether complete-case
analysis or multiple imputation was used. Table 2 (complete-case
analysis) shows that participants who associated chocolate cake
with celebration gained less weight (M = 0.36 kg) than did participants who associated chocolate cake with guilt (M = 2.36 kg).
Although there were a few signicant correlations between
baseline eating measures (behaviours, evaluation and TPB variables) and follow-up weight change, only the correlation between

Participants who associated chocolate cake with guilt or celebration did not differ in the total amount of weight they wanted
to lose. However, participants who responded with guilt had a
more ambitious 3 month goal compared to participants who
responded with celebration (see Table 2). Participants who associated chocolate cake with guilt also rated their weight-loss goal as
more important compared to participants who responded with celebration (Table 2). The groups did not differ in condence.
At follow-up 3 months later, participants weighed on average
the same as they did at baseline (M = 0.11 kg weight gain;
SD = 2.80; range: 7 kg weight-loss10 kg weight-gain). However,
participants who associated chocolate cake with celebration were
on average more successful in losing weight (M = 0.43 weight-loss)
than were participants who associated chocolate cake with guilt
(M = 1.25 weight-gain) (complete case-analysis, see Table 2). The
main effect for default association was signicant regardless of
whether complete-case analysis or multiple imputation was used.
On average participants in the weight-loss sample reported that
they had not tried very hard to reach their goal (M = 4.06 on a scale
from 1 to 10). Although participants who associated chocolate cake
with celebration lost more weight than did participants who associated chocolate cake with guilt, no differences were found
between the two groups with respect to effort (Table 2) and
amount of effort was unrelated to weight change (r = .12, ns). Neither eating behaviours, TPB variables, nor any of the other weightloss variables assessed at baseline were signicantly correlated
with 3 month weight-loss (all rs < .14, ns).

R.G. Kuijer, J.A. Boyce / Appetite 74 (2014) 4854

Discussion
With the current increase in obesity and associated health risks,
many countries (including New Zealand) have launched (media)
campaigns and programmes to increase healthy dietary habits
and prevent overweight. An unintended side effect of these campaigns and programmes may be that they fuel feelings of guilt
and worry about food (cf. Rozin et al., 1999, 2003). For example,
a study by Harting, van Assema, and de Vries (2006) showed that
after health counselling by a health advisor about high fat consumption, smoking and physical inactivity, one third of the participants reported an increased guilty conscience. But is guilt good or
bad?
In our study, a little over one quarter of the sample associated
chocolate cake more with guilt than with celebration. This percentage was comparable to that found by Rozin et al. (2003) in a
student sample. As expected, we found that participants with a
weight-loss goal were more likely to associate chocolate cake with
guilt than were participants without a weight-loss goal (e.g., Cartwright & Stritzke, 2008). However, even in a subsample of participants who wanted to lose weight, the majority (63%) associated
chocolate cake with celebration rather than guilt.
The current study did not nd any evidence for adaptive or
motivational properties of guilt. Those who associated chocolate
cake with guilt did not report more positive attitudes toward
healthy eating or stronger intentions to eat healthy in the future
than did those who associated chocolate cake with celebration. Instead, they reported lower levels of perceived behavioural control
over eating (i.e., lower levels of self-efcacy) and were less successful at maintaining their weight over an 18 month period (whilst
taking into account important covariates such as BMI, age and education). Moreover, those with an active weight-loss goal who associated chocolate cake with guilt were less successful at losing
weight over a 3 month period (they actually gained weight) compared to those who associated chocolate cake with celebration.
To our knowledge this is the rst study that has examined the link
between a guilty food-attitude and actual weight change over time
in a healthy sample. Our ndings are in line with Rozins suggestions (Rozin et al., 1999, 2003) that worry, concern and guilt about
food are counterproductive.
Although our ndings support Rozins theory (Rozin et al., 1999,
2003), they do not shed any light on possible underlying mechanisms. Several researchers have suggested that guilt leads to feelings of helplessness and loss of control, and it is those feelings
that result in maladaptive outcomes (cf. Tangney et al., 2007). Guilt
may also signal a conict between approach and avoidance motivations, whereas celebration does not. Just as trying to suppress
an unwanted thought (do not think of a white bear) often results
in this thought actually becoming more prevalent and intrusive
(Wegner, 1994), trying to avoid forbidden foods makes them more
desirable and tends to result in cravings (Rodgers et al., 2011; Rogers & Smit, 2000) and hence less control. In the current study, those
who associated chocolate cake with guilt indeed reported lower
levels of perceived behavioural control over healthy eating, and
both lower levels of perceived behavioural control and associating
chocolate cake with guilt were related to less successful weight
maintenance. Future research should examine a formal mediation
model where each variable in the model is assessed at a different
time point.
There were no other variables assessed at baseline that could
explain the link between associating chocolate cake with celebration and more successful weight maintenance. Participants associating chocolate cake with celebration did not have healthier eating
intentions or attitudes, nor did they try harder to lose weight. And
although they did evaluate their eating as healthier at baseline, this

53

did not affect their weight over time. In line with research showing
a link between chocolate-related guilt and higher body dissatisfaction and drive for thinness (Rodgers et al., 2011), we found that
those with a weight-loss goal associating chocolate cake with guilt
(vs. celebration) had a more ambitious 3 month weight-loss goal
and rated their goal as more important. Perhaps having a more
ambitious goal set these participants up for failure and disappointment (Foster, Wadden, Vogt, & Brewer, 1997). However, neither
the amount of weight participants wanted to lose nor importance
of their weight goal (or condence) was related to 3 month weightchange.
The nding that associating chocolate cake with celebration
was related to more successful weight maintenance ts nicely with
a recent focus in the broader psychological literature on the benecial effects of positive affective states on peoples outcomes in life
(see Lyubomirsky, King, & Diener, 2005). According to Fredericksons (2001) broaden-and-build theory, positive mood works as
a resource, making challenges such as self-control dilemmas easier
to deal with. There is also some evidence from laboratory studies
that positive mood may enhance self-control and decrease unhealthy food intake (Fedorikhin & Patrick, 2010; Turner, Luszczynska, Warner, & Schwarzer, 2010; Winterich & Haws, 2011),
although not all research shows such positive effects (e.g., Evers,
Adriaanse, de Ridder, & de Witt Huberts, 2013). Conversely, Lindeman and Stark (2000) found that especially dieters who rated
ideological issues as important and pleasure as unimportant when
choosing food, showed more signs of eating disorders and lower
psychological well-being compared to other dieters.
The way in which chocolate-related guilt was measured in the
present study deserves some attention. An advantage of measuring
it as a default association is that it taps into the most salient way a
person thinks about a certain issue (see Rozin et al., 1999, 2003).
Like free associations, such forced-choice items demand less
deliberate considerations and may capture more spontaneous
and automatic representations that are not accessible with explicit
questionnaires. A disadvantage of our current measure is that we
have no information about the temporal aspect of those feelings
of guilt (or celebration), that is, we do not know whether those
people who responded with guilt did so because they usually feel
guilty before or after eating chocolate cake, or a mixture of both.
Anticipated guilt has more potential to be motivating (as there
are opportunities to change ones behaviour) than guilt that is
experienced after a transgression or violation (Giner-Sorolla,
2001). A related issue is that we only used one item to assess the
default association in relation to chocolate which is a limitation.
In addition, although chocolate is the prototypical example of a forbidden food (and indeed the majority of studies in the eating domain that have examined eating related guilt have looked at
chocolate), it would be important to consider default associations
toward other food items in future research.
The current study has some limitations that should be addressed. First, the study relied on self-report measures for weight,
and height. Although self-reported and measured BMI are highly
correlated, people tend to slightly overestimate their height and
underestimate their weight (e.g., Spencer, Appleby, Davey & Key,
2002). However, if some people indeed underestimated their
weight then it is likely they will have done so at each measurement
time, hence probably not inuencing the weight change measure.
We also used a self-report measure to assess eating behaviours
over the past 2 weeks. To improve our understanding of exactly
how guilt affects eating behaviour, it is essential that future research uses more objective behavioural measures such as observation of eating behaviour, or diary methods. Second, we controlled
for sex in our analyses where appropriate, but future research
should systematically examine sex as a moderator variable. Our

54

R.G. Kuijer, J.A. Boyce / Appetite 74 (2014) 4854

sample did not have enough male participants for such analyses. In
line with other research we found that women were more likely to
report chocolate related guilt than were men (Cramer & Hartleib,
2001; Mller et al., 2008). However, it is also possible that women
and men are differently affected by guilt and future research
should examine this possibility. Third, as participants were not
originally recruited for a multi-measurement study, there was
considerable drop-out between baseline and follow-ups. Younger
participants, those associating chocolate cake with guilt and those
with less healthy eating behaviours and attitudes were less likely
to complete the follow-up measures. To address the issue of
non-response bias, multiple imputation was used in addition to
complete-case analysis. Both methods yielded the same results
suggesting that the differences between completers and drop-outs
did not bias the results.
To conclude, the current study did not nd evidence for adaptive properties of guilt. Increasing peoples guilty conscience about
food and eating through media campaigns and programmes about
healthy nutrition with the view to initiate or motivate behaviour
change seems ill advised. Enjoyment of food and eating is essential
to peoples well-being and the current study shows that people
who associate a forbidden food with celebration and view it as a
treat that can be enjoyed do better in terms of weight management. In education messages about dietary recommendations
enjoyment of food and eating should receive more attention than
it has in the past.
References
Ajzen, I. (1991). The theory of planned behaviour. Organizational Behavior and
Human Decision Processes, 50, 179211.
Baker, C. W., Little, T. D., & Brownell, K. D. (2003). Predicting adolescent eating and
activity behaviors. The role of social norms and personal agency. Health
Psychology, 22, 189198.
Baumeister, R. F., & Heatherton, T. (1996). Self-regulation-failure. An overview.
Psychological Inquiry, 7, 115.
Cartwright, F., & Stritzke, W. G. K. (2008). A multidimensional ambivalence model of
chocolate craving. Construct validity and associations with chocolate
consumption and disordered eating. Eating Behaviors, 9, 112.
Conner, M., Norman, P., & Bell, R. (2002). The theory of planned behaviour and
healthy eating. Health Psychology, 21, 194201.
Conradt, M., Dierk, J. M., Schlumberger, P., Rauh, E., Hebebrand, J., & Rief, W. (2008).
Who copes well? Obesity related coping and its associations with shame, guilt
and weight-loss. Journal of Clinical Psychology, 64, 11291144.
Cramer, K. M., & Hartleib, M. (2001). The attitudes to chocolate questionnaire. A
psychometric evaluation. Personality and Individual Differences, 31, 931942.
De Witt Huberts, J. C., Evers, C., & de Ridder, D. T. D. (2013). Double trouble.
Restrained eaters do not eat less and feel worse. Psychology & Health, 28,
686700.
Dewberry, C., & Ussher, J. M. (1994). Restraint and perception of body weight among
British adults. The Journal of Social Psychology, 134, 609619.
Evers, C., Adriaanse, M., de Ridder, D. T. D., & de Witt Huberts, J. C. (2013). Good
mood food. Positive emotion as a neglected trigger for food intake. Appetite, 68,
17.
Fedorikhin, A., & Patrick, V. M. (2010). Positive mood and resistance to temptation.
The interfering inuence of elevated arousal. Journal of Consumer Research, 37,
698711.
Foster, G., Wadden, T., Vogt, R., & Brewer, G. (1997). What is a reasonable weight
loss? Patients expectations and evaluations of obesity treatment outcomes.
Journal of Consulting and Clinical Psychology, 65, 7985.
Frederickson, B. L. (2001). The role of positive emotions in positive psychology. The
broaden-and-build theory of positive emotions. American Psychologist, 56,
218226.

Giner-Sorolla, R. (2001). Guilty pleasures and grim necessities. Affective attitudes in

dilemmas of self-control. Journal of Personality and Social Psychology, 80,
206221.
Harting, J., van Assema, P., & de Vries, N. K. (2006). Patients opinions on health
counselling in the Hartslag Limburg cardiovascular prevention project.
Perceived quality, satisfaction, and normative concerns. Patient, Education and
Counselling, 61, 142151.
Herman, C. P., & Polivy, J. (1984). A boundary model for the regulation of eating.
Psychiatric Annals, 13, 918927.
King, G. A., Herman, P., & Polivy, J. (1987). Food perception in dieters and nondieters. Appetite, 8, 147158.
Kuijer, R. G., & Boyce, J. A. (2012). Emotional eating and its effect on eating
behaviour after a natural disaster. Appetite, 58, 936939.
Lindeman, M., & Stark, K. (2000). Loss of pleasure, ideological food choice reasons
and eating pathology. Appetite, 35, 263268.
Lyubomirsky, S., King, L., & Diener, E. (2005). The benets of frequent positive affect.
Does happiness lead to success? Psychological Bulletin, 131, 803855.
Macht, M., & Dettmer, D. (2006). Everyday mood and emotions and eating a
chocolate bar or an apple. Appetite, 46, 332336.
Ministry of Health (2008). A portrait of health. Key results of the 2006/07 New Zealand
health survey. Wellington: Ministry of Health.
Mller, J., Dettmer, D., & Macht, M. (2008). The attitudes to chocolate questionnaire.
Psychometric properties and relationship to dimensions of eating. Appetite, 50,
499505.
OECD Health Data (2011). Obesity update 2012. <http://www.oecd.org/health/
49716427.pdf>.
Rodgers, R. F., Stritzke, W. G. K., Bui, E., Franko, D. L., & Chabrol, H. (2011). Evaluation
of the French version of the orientation towards chocolate questionnaire.
Chocolate-related guilt and ambivalence are associated with overweight and
disordered eating. Eating Behaviors, 12, 254260.
Rogers, P. J., & Smit, H. J. (2000). Food craving and food addiction. A critical review
of the evidence from a biopsychological perspective. Pharmacology Biochemistry
and Behavior, 66, 314.
Rozin, P., Bauer, R., & Catanese, D. (2003). Food and life, pleasure and worry, among
American college students. Gender differences and regional similarities. Journal
of Personality and Social Psychology, 85, 132141.
Rozin, P., Fischler, C., Imada, S., Sarubin, A., & Wrzesniewski, A. (1999). Attitudes to
food and the role of food in life in the U.S.A., Japan, Flemish Belgium and France.
Possible implications for the diet-health debate. Appetite, 33, 163180.
Rozin, P., Kurzer, N., & Cohen, A. B. (2002). Free associations to food. The effects of
gender, generation and culture. Journal of Research in Personality, 36, 419441.
Sassaroli, S., Bertelli, S., Decoppi, M., Crosina, M., Milos, G., & Ruggiero, G. M. (2005).
Worry and eating disorders. A psychopathological association. Eating Behaviors,
6, 301307.
Schafer, J. L., & Graham, J. W. (2002). Missing data. Our view of the state of the art.
Psychological Methods, 7, 147177.
Spencer, E. A., Appleby, P. N., Davey, G. K., & Key, T. J. (2002). Validity of selfreported height and weight in 4808 EPIC-Oxford participants. Public Health
Nutrition, 5, 561565.
Stice, E. (1998). Relations of restraint and negative affect to bulimic pathology. A
longitudinal test of three competing models. International Journal of Eating
Disorders, 23, 243260.
Swaim, R. C., Wayman, J. C., & Chen, J. (2004). Alcohol use among Mexican American
and non-Hispanic white 7th12th-grade students in Southwestern United
States. Journal of Child and Adolescent Substance Abuse, 14, 118.
Tangney, J. P., Stuewig, J., & Mashek, D. J. (2007). Moral emotions and moral
behaviour. Annual Review of Psychology, 58, 345372.
Turner, S. A., Luszczynska, A., Warner, L., & Schwarzer, R. (2010). Emotional eating
and uncontrolled eating styles and chocolate chip cookie consumption. A
controlled trial of the effects of positive mood enhancement. Appetite, 54,
143149.
Wegner, D. M. (1994). Ironic processes of mental control. Psychological Review, 101,
3452.
Winterich, K. P., & Haws, K. L. (2011). Helpful hopefulness. The effect of future
positive emotions on consumption. Journal of Consumer Research, 38, 505524.
Witt, A. A., Katterman, S. N., & Lowe, M. R. (2013). Assessing the three types of
dieting in the three-factor model of dieting. The dieting and weight history
questionnaire. Appetite, 63, 2430.