Appetite (2002) 38, 39±44

1
doi:10.1006/appe.2001.0445, available online at http://www.idealibrary.com on

Original Article

Rigid vs. flexible dieting: association with

eating disorder symptoms in nonobese women
Tiffany M. Stewart, Donald A.Williamson and Marney A.White
Louisiana State University and Pennington Biomedical Research Center, Baton Rouge, LA, U.S.A.

(Received1August 2001, finalrevision16 November 2001, accepted in revised form 20 December 2001)

The correlates of rigid and flexible dieting were examined in a sample of 188 nonobese women recruited from the
community and from a university. The primary aim of the study was to test the hypothesis that women who utilize
rigid versus flexible dieting strategies to prevent weight gain report more eating disorder symptoms and higher body
mass index (BMI) in comparison to women who utilize flexible dieting strategies. The study sample included women
who were underweight (29%), normal weight (52%), and overweight (19%). None of the women were obese, as
defined by BMI > 30. Participants were administered a questionnaire that measures Rigid Control and Flexible
Control of eating. Body weight and height were measured and measures of eating disorder symptoms and mood
disturbances were administered. Our results indicated that BMI was significantly correlated with rigid dieting and
flexible dieting. BMI was controlled statistically in other analyses. The study found that individuals who engage in
rigid dieting strategies reported symptoms of an eating disorder, mood disturbances, and excessive concern with
body size/shape. In contrast, flexible dieting strategies were not highly associated with BMI, eating disorder
symptoms, mood disturbances, or concerns with body size. Since this was a cross sectional study, causality of eating
disorder symptoms could not be addressed. These findings replicate and extend the findings of earlier studies. These
findings suggest that rigid dieting strategies, but not flexible dieting strategies, are associated with eating disorder
symptoms and higher BMI in nonobese women.
# 2002 Elsevier Science Ltd

Introduction he called ``rigid control'' and ``flexible control'' of

The term dietary restraint refers to a person's intent to
restrict dietary intake to control body weight (Herman &
Mack, 1975). Dietary restraint has been hypothesized
to play an important role in determining binge eating
associated with eating disorders (Polivy & Herman,
1985). The Three Factor Eating Questionnaire (TFEQ),
developed by Stunkard and Messick (1985), measures
dietary restraint, disinhibition (overeating), and per-
ceived hunger. The dietary restraint scale of the TFEQ
measures cognitive control of eating behaviour. In a
subsequent study of the TFEQ, Westenhoefer (1991)
found evidence that the construct of dietary restraint
could be separated into two dieting strategies, that
Address correspondence to: Donald A. Williamson, Ph.D.,
Pennington Biomedical Research Center, 6400 Perkins Road,
Baton Rouge, LA 70808, U.S.A.
E-mail: Williada@pbrc.edu
eating behaviour. Rigid control was characterized by
dichotomous ``all or nothing'' approach to eating,
weight, and dieting. Rigid control was associated with
a higher body mass index, more frequent and more
severe binge eating episodes, and higher scores of
disinhibition (overeating). In contrast, flexible control
was characterized by a more graduated approach to
eating, weight, and dieting. Flexible control was
associated with lower body mass index, less frequent
and less severe binge eating or overeating, and a higher
probability of successful weight loss during a one-year
weight loss program. Subsequent research (McGuire et
al., Shearin et al., 1994; Smith et al., 1998; Williamson
et al., 1995) provided partial support for the initial
findings of Westenhoefer (1991). A key difference in
the findings of these studies was that some investiga-
tions found that flexible control was associated with
lower BMI and others did not.
Westenhoefer, Stunkard and Pudel (1999) added
items to the original scales and validated two revised

0195±6663/02/010039+06 $35.00/0 # 2002 Elsevier Science Ltd

40 T. M. Stewart et al.
scales called the Flexible and Rigid Control dimensions
of dietary restraint. The end product was a 12- item scale
measuring Rigid Control and a 16-item scale measuring
Flexible Control. The validation studies for these scales
used correlations of the two scales with body mass index,
the disinhibition scale of the TFEQ, and self-reported
eating disorder symptoms. The findings of the 1999
paper supported the conclusions of Westenhoefer (1991)
that rigid control was positively correlated with higher
BMI and greater eating disorder symptoms; whereas
flexible control was negatively correlated with BMI and
with eating disorder symptoms. Westenhoefer et al.
(1999) studied German men and women with a wide
range of BMI.
In the study reported in this paper, adult women with
body weights ranging from underweight to overweight
(BMI < 30) were selected for study. This selection of
nonobese women is important for the following rea-
sons: (1) the adverse effects of dieting are of greatest
concern for nonobese women (National Task Force
on the Treatment and Prevention of Obesity, 2000);
(2) the relevance of the hypothesis that certain dieting
strategies are harmful is strongest in nonobese women,
since these are the people most vulnerable for develop-
ing anorexia and bulimia nervosa; (3) most of the pre-
vious tests of this hypothesis investigated samples of
predominantly obese men and women that included
some obese participants. The primary aim of the study
was to test the hypothesis that Rigid Control of eating,
but not Flexible Control of eating, is associated with
ED symptoms, mood disturbance, and higher BMI.
A secondary aim was to test the hypothesis that Flexible
Control of eating is negatively correlated with eating
disorder symptoms and BMI.
Method
Subjects
A sample of 188 female participants was recruited
from undergraduate courses at a large university and
from the community residents. Inclusion criteria were:
(1) female gender, (2) between the ages of 18 and 65
years, and (3) BMI < 30. Approximately one-half of
the sample was recruited from undergraduate classes,
and the other half from the community. With regard to
weight status, 30
3% of the participants had a Body
Mass Index (BMI) under 20 (underweight); 54
3% of
the participants had a BMI between 20 and 25 (normal
weight), and 15
4% had a BMI between 25 and 30
(overweight) (Bray, 1998). The sample contained
primarily Caucasian women (89
3%), with some
African American women (5
3%), Hispanic women
(2
7%), and women of other racial groups (2
7%). The
mean age of the sample was 22.6 (SD ˆ 6
8). The mean
height (m) and weight (kg) of the overall sample was
1
7 (SD ˆ 0
07) and 59
0 (SD ˆ 7
7), respectively. The
mean BMI of the sample was 21
7 (SD ˆ 2
7).
Assessment procedures
Rigid vs. Flexible Dieting Scale (Westenhoefer et al.,
1999). This self-report inventory has 12 questions that
measure Rigid Control of eating and 14 questions that
measure Flexible Control of eating. The Rigid and
Flexible Control scales were an extension of earlier
research by Westenhoefer (1991) pertaining to the con-
struct of dietary restraint. The two scales were found
to have satisfactory reliability and validity in a series
of studies reported by Westenhoefer et al. (1999). In
this study, the psychological and behavioural corre-
lates of the Rigid and Flexible Control scales were
investigated. The psychological measures that were
tested as correlates of Rigid and Flexible Control
are described below. Each measure was selected to
assess some aspect of the symptoms that are broadly
associated with eating disorders (Williamson, 1990).
Body Mass Index. Self-reported measurements of
height (m) and weight (kg) were obtained from partici-
pants. Self-reported weights have been shown to be
correlated with weight measured in the clinic (Bowman
& DeLucia, 1992). From these values, BMI was calcu-
lated using the formula: body weight (kg)/height (m2).
Beck Depression Inventory II (BDI-II; Beck, Brown &
Steer, 1996). The BDI is a 21-item self-report ques-
tionnaire measure that was utilized to assess the cogni-
tive, behavioural, affective, and somatic symptoms of
depression in the participants. The BDI has good
internal consistency and is correlated with other
measures of depression (Beck et al., 1996).
Eating Attitudes Test (EAT; Garner & Garfinkel,
1979). The EAT is a 40-item self-report inventory that
assesses behaviours and attitudes associated with
anorexia nervosa (Garner & Garfinkel, 1979). The reli-
ability and validity of the EAT as a measure of eating
disorder symptoms has been supported by many studies
(Garner & Garfinkel, 1979; Williamson, 1991).
Perceived Weight Cycling (PWC; Friedman Schwartz
& Brownell, 1998). The perception that a person has a
history of gaining and losing weight was measured
using a 3 item scale. This scale was found to be corre-
lated with psychological problems associated with
weight cycling (Friedman et al., 1998).
Body Shape Questionnaire (BSQ; Cooper et al., 1987).
The BSQ measures concerns with body shape, in par-
ticular, the experience of ``feeling fat'' (Cooper et al.,
1987). The BSQ is correlated with measures of body

dissatisfaction and has been used to define subthreshold
eating disorders in nonobese women (Williamson et al.,
2000).
Three Factor Eating Questionnaire (TFEQ; Stunkard
& Messick, 1985). The TFEQ (also called the Eating
Inventory) is a 51-item questionnaire that measures
three constructs: dietary restraint, disinhibition, and
perceived hunger. The Dietary Restraint scale mea-
sures self-reported restrictive eating and dieting beha-
viour and many of the items of the Rigid and Flexible
Control scales are included in the Dietary Restraint
scale. Because of this item overlap, only scores from
the disinhibition (TFEQ-D) and perceived hunger
(TFEQ-PH) scales were included in the statistical
analysis.
State-Trait Anxiety Inventory (STAI; Spielberger
et al., 1970). The STAI is a self-report measure desig-
ned to measure situational (STAI-State) and generali-
zed anxiety (STAI-Trait). The reliability and validity of
the STAI have been established as a measure of anxiety
in a wide variety of studies (Spielberger et al., 1970;
Williamson, 1990).
Procedure
From an initial sample of 200 women, 12 were
excluded; 8 for BMI>30, and 4 for incomplete data,
leaving a study sample of 188 women. Women who
were recruited from the student population were given
extra credit in undergraduate classes. All other
participants were volunteers who did not receive any
incentive to participate in the study. All volunteers
signed consent to participate in the study. The
participants completed the questionnaire packets,
which included questions about self-reported height
and weight, age, race/ethnicity, and marital status.
Results
The Rigid and Flexible dieting scales were highly
correlated (r(186) ˆ 0
70, p < 0
01). Table 1 sum-
marizes the correlations among the Rigid and Flexible
Control scales with BMI and the psychological
variables as assessed by the BDI-II, EAT, PWC,
BSQ, TFEQ-D, TFEQ-PH, STAI-State, and STAI-
Trait scales. Flexible Control was positively correlated
with BMI, EAT, PWC, and BSQ scores. Rigid Control
was positively correlated with BMI, BDI-II, EAT,
PWC, BSQ, TFEQ-D, TFEQ-PH, STAI-State anxiety,
and STAI-Trait anxiety scores. Correlations between
psychological variables and Rigid Control were con-
sistently higher than correlations between psychologi-
cal variables and Flexible Control scores.
Rigid vs. flexible dieting 41
Table 1. Correlation analyses
Flexible control Rigid control
BMI 0
22* 0
32*
BDI-II 0
18 0
26*
EAT 0
48* 0
66*
PWC 0
29* 0
55*
BSQ 0
44* 0
70*
TFEQ-D 0
18 0
49*
TFEQ-PH ÿ0
05 0
19*
STAI-State 0
13 0
26*
STAI-Trait 0
15 0
30*
Abbreviations: BMI ˆ Body Mass Index, BDI-II ˆ Beck Depression
Inventory, EAT ˆ Eating Attitudes Test, PWC ˆ Perceived Weight
Cycling Measure, BSQ ˆ Body Shape Questionnaire, TFEQ ˆ The
Three Factor Eating Questionnaire, D ˆ Disinhibition, PH ˆ Per-
ceived Hunger, STAI-State ˆ State-Trait Anxiety Inventory State
anxiety scale, STAI-Trait ˆ State-Trait Anxiety Inventory Trait
anxiety scale.
*Correlation is significant at the 0
01 level (2-tailed). Degrees of
freedom for all correlations were 186.
Since the Rigid and Flexible Control scales were
highly correlated, we were interested in the psycho-
logical and behavioural characteristics of groups of
participants that scored relatively high or low on the two
scales, while statistically controlling for variation
associated with BMI (since BMI was correlated with
both Rigid and Flexible Control scales). To accomplish
this end, participants were categorized into dieting
strategy groups using a median split of scores on the
Rigid (median ˆ 7) and Flexible (median ˆ 5) Control
scales. Participants scoring below or at the median
were classified as ``low'' while those scoring above the
median were classified as ``high''. Using this procedure,
approximately 10
1% (N ˆ 19) of the sample could
be classified as ``Predominantly'' Rigid dieters and
13
3% (N ˆ 25) as ``Predominantly'' Flexible dieters.
For example, a Predominantly Rigid dieter scored above
the median on the Rigid Control scale, but below the
median on the Flexible Control scale. The remaining
portion of the sample was also classified according to
median splits on both scales: 42
0% (N ˆ 79) were
classified as Low Dieters on both scales, and 34
6%
(N ˆ 65) were classified as High Dieters on both scales.
A 2 (flexible category)  2 (rigid category) analysis of
variance (ANOVA) was used to test whether BMI dif-
fered across dieting strategy groups. There was no main
effect for Flexible control, F(1,184) ˆ 2
406, p > 0
10.
Also, there was no significant interaction of Rigid and
Flexible control, F(1,184) ˆ 0
000, p > 0
99. There was a
significant main effect for Rigid Control, F(1,184) ˆ
4
179, p < 0
05. The mean BMI for the Low Rigid
Control group was 21
14 (SD ˆ 2
73) and the mean
BMI for the High Rigid Control group was 22
46
(SD ˆ 2
55).
42 T. M. Stewart et al.

Table 2. Estimated marginal means (means adjusted for BMI as covariate)

Dependent variable Rigid median category Mean SE F(1, 182) p
BDI-II Low (7) 6
616 0
958 8
842 0
003
High (8) 10
954 1
089
EAT Low (7) 9
987 1
158 35
336 0
001
High (8) 20
473 1
317
PCW Low (7) 5
398 0
291 23
003 0
001
High (8) 7
522 0
331
BSQ Low (7) 80
865 3
318 48
849 0
001
High (8) 116
192 3
773
TFEQ-D Low (7) 5
303 0
422 23
551 0
001
High (8) 8
42 0
479
TFEQ-PH Low (7) 5
594 0
379 6
083 0
015
High (8) 7
017 0
431
STAI-State Low (7) 48
809 1
213 6
38 0
012
High (8) 53
477 1
38
STAI-Trait Low (7) 51
257 1
334 4
256 0
041
High (8) 55
451 1
517
Abbreviations: BDI-II ˆ Beck Depression Inventory, EAT ˆ Eating Attitudes Test, PCW ˆ Perceived Weight Cycling
Measure, BSQ ˆ Body Shape Questionnaire, TFEQ ˆ Three Factor Eating Questionnaire, D ˆ Disinhibition,
PH ˆ Perceived Hunger, STAI-State ˆ State-Trait Anxiety Inventory State anxiety scale, STAI-Trait ˆ State-Trait
Anxiety Inventory Trait anxiety scale.

Ordinal regression analyses were conducted on the
categorical demographic variables and revealed no
differences in the proportion of predominantly rigid
versus flexible dieting strategies across racial/ethnic
groups, 2(10) ˆ 5
04, p > 0
05, or marital status groups
2(4) ˆ 0
74, p > 0
05.
Group differences on each of the psychological
variables were first tested using MANCOVA, with a
2 (High/Low Rigid)  2(High/Low Flexible) data ana-
lytic plan. BMI was used as a covariate since BMI was
significantly correlated with both Flexible and Rigid
Control scores. BMI was found to be a significant co-
variate F(8, 175) ˆ 5
203, p < 0.01. There was a signifi-
cant main effect for Rigid Control, F(8, 175) ˆ 8
396,
p < 0
01, but the main effect for Flexible Control
F(8, 175) ˆ 1
662, p > 0
10, and the Flexible  Rigid
interaction F(8, 175) ˆ 0
886, p > 0
5 were not statisti-
cally significant. Follow-up analyses using ANCOVAs
for each of the psychological variables are reported in
Table 2. High and Low Rigid Control groups differed
on each of the psychological variables. The proportion
of predominantly Rigid versus Flexible dieters did
not differ as a function of the BMI category (i.e., under
weight, normal weight, and overweight), 2 (2, N ˆ
44) ˆ 3
97, p > 0
10. This multivariate data analytic
strategy was selected because the study hypothesis was
that rigid, but not flexible dieting, would be associated
with eating disorder symptoms, mood disturbances, and
higher BMI. This hypothesis called for the use of a
multivariate omnibus test (Huberty & Morris, 1989).
We also tested the hypothesis using a series of multi-
variate regression analyses using Rigid Control, Flexible
Control, and BMI as continuous variables, controlling
for experiment-wise error. These analyses resulted in
identical conclusions to the MANCOVA approach.
Discussion
This study supported the primary hypothesis that
individuals engaging in rigid dieting strategies were
more likely to report symptoms of an eating disorder,
including mood disturbances and higher anxiety. The
findings of the study did not support the secondary
hypothesis that higher Flexible Control scores were
associated with lower BMI and fewer symptoms of
eating disorders. It should be noted that the sample for
this study was 100% women and that there were no
obese participants in the study sample. A high
correlation between the Rigid and Flexible Control
scales was found. The only other study of the revised
Rigid and Flexible Control scales (Westenhoefer et al.,
1999) did not report the correlation between the two
scales. The finding of a strong correlation between the
two scales suggested that the women of this sample did
not often rely upon a single dieting strategy to control
body weight. To investigate this possibility, we divided
the sample into predominantly rigid versus flexible
dieters using a median split of the scores on each scale.
Only 23% of the sample could be classified as
Predominately Rigid or Flexible dieters. A larger pro-
portion (35%) reported using both dieting strategies.
A higher body mass index was associated with rigid

dieting strategies, but not flexible dieting strategies.
In the comparison of rigid versus flexible dieting groups
on psychological variables, differences in BMI were
controlled statistically. Group differences attributed
to Rigid Control were not diminished by covariance
of BMI. This pattern of results indicates a higher
association of eating disorder symptoms and body
image disturbance with rigid dieting strategies
than with flexible dieting strategies, when covari-
ance with BMI and other variables was controlled
statistically.
A limitation of the study is that the findings of
the study were primarily based on self-report data.
In addition, the sample consisted of primarily healthy
young people. This could have limited the occurrence
of eating disorder symptoms in the sample, biasing
the results. Another limitation is that the conclusions
of the study were based upon correlational findings,
which does not imply causation of the relationship
between dieting strategies and behaviour. Thus, the
directionality of the association between rigid dieting
and eating pathologies is not known; it could be that
individuals with greater concern with body size and
shape and eating-disordered behaviours are more
likely to engage in rigid dieting strategies. Since this
was a cross sectional study, causal associations
among the variables were not addressed. There is a
need for longitudinal studies to test for causal asso-
ciations between Rigid versus Flexible Control of
eating and the development of eating disorder
symptoms and/or increased body weight.
In summary, this is the first study to report a strong
correlation between the Rigid and Flexible Control
scales. This is the second study that reports on the new
Rigid and Flexible Dieting scales. The findings of this
study partially replicated and extended the findings of
Westenhoefer et al. (1999). Since Rigid and Flexible
Control scores were highly correlated and since both
scales were positively correlated with BMI, it was
imperative to control for covariation among the vari-
ables, which was accomplished using a multivariate
statistical approach. The finding that Rigid Control
scores were associated with eating disorder symptoms,
broadly defined, replicated and extended the basic
conclusions of most other studies (e.g. Smith et al., 1999;
Westenhoefer, 1991; Westenhoefer et al., 1999) that has
implicated that rigid, but not flexible dieting strategies
are associated with eating disorder symptoms such as
binge eating. However, this study and the recently
published paper of McGuire et al. (2001) failed to find a
negative correlation between Flexible Control and BMI.
Other studies, however, have reported a negative cor-
relation between Flexible Control and BMI (Smith et al.,
1999; Westenhoefer et al., 1999; Williamson et al. 1995).
Rigid vs. flexible dieting 43
Thus, the results of these studies have consistently
supported the positive relationship between Rigid
Control and eating disorder symptoms and BMI, but the
negative correlation between Flexible Control and BMI
and eating disorder symptoms has not been consistently
supported.
References
Beck, A. T., Brown, G. K. & Steer, R. A. (1996). The Beck
Depression Inventory II. San Antonio, TX, Psychological
Corporation.
Bowman, R. L. & DeLucia, J. L. (1992). Accuracy of self-
reported weight: A meta-analysis. Behaviour Therapy
23, 637±655.
Bray, G. A. (1998). Classification and evaluation of the over-
weight patient. In G. Bray, C. Bouchard, and W. James
(Eds.) Handbook of Obesity. pp. 831±854. New York:
Marcel Dekker, Inc.
Cooper, P. J., Taylor, B. S., Cooper, Z. & Fairburn, M. D.
(1987). The Development and Validation of the Body
Shape Questionnaire. International Journal of Eating
Disorders 6(4), 485±494.
Friedman, M. A., Schwartz, M. B. & Brownell, K. D. (1998).
Differential relation of psychosocial functioning with
the history and experience of weight cycling. Journal of
Consulting and Clinical Psychology 4, 646±650.
Garner, D. M. & Garfinkel, P. E. (1979). The Eating Atti-
tudes Test: an index of symptoms of anorexia nervosa.
Psychological Medicine 9, 273±279.
Herman, C. P. & Mack, D. (1975). Restrained and unre-
strained eating. Journal of Personality 43, 647±660.
Huberty, C. J. & Morris, J. D. (1989). Multivariate analysis
versus multiple univariate analyses. Psychological Bulletin
105, 302±308.
McGuire, M. T., Jeffery, R. W., French, S. A. & Hannan, P. J.
(2001). The relationship between restraint and weight and
weight-related behaviours among individuals in a com-
munity weight gain prevention trial. International Journal
of Obesity 25, 574±580.
National Task Force on the Prevention and Treatment of
Obesity. (2000). Archives of Internal Medicine 160,
2581±2588.
Polivy, J. & Herman, C. P. (1985). Dieting and binging:
A causal analysis. American Psychologist 40, 193±201.
Shearin, E. N., Russ, M. J., Hull, J. W. & Clarkin, J. F.
(1994). Construct validity of the Three-Factor Eating
Questionnaire: Flexible and rigid control subscales. Inter-
national Journal of Eating Disorders 16(2), 187±198.
Smith, C. F., Williamson, D. A., Bray, G. A. & Ryan, D. H.
(1999). Flexible vs. Rigid Dieting Strategies: Relation-
ship with Adverse Behavioural Outcomes. Appetite 32,
295±305.
Spielberger, C. D., Gorsuch, R. L. & Lushene, R. E. (1970).
Manual for the State Trait Anxiety Inventory. Palo Alto,
CA: Consulting Psychologists Press.
Stunkard, A. J. & Messick, S. (1985). The three-factor eating
questionnaire to measure dietary restraint, disinhibition,
and hunger. Journal of Psychosomatic Research 29, 71±83.
Westenhoefer, J. (1991). Dietary restraint and disinhibition:
Is restraint a homogeneous construct? Appetite 16,
45±55.
44 T. M. Stewart et al.

Westenhoefer, J., Stunkard, A. & Pudel, V. (1999). Valida- Williamson, D. A., Perrin, L., Blouin, D. C. & Barbin, J. M.
tion of the flexible and rigid control dimensions (2000). Cognitive bias in eating disorders: Interpretation
of dietary restraint. International Journal of Eating of ambiguous body-related information. Eating and
Disorders 26, 53±64. Weight Disorders 5, 143±151.
Williamson, D. A. (1990). Assessment of eating disorders:
Obesity, anorexia, and bulimia nervosa. New York:
Pergamon Press.