Construct validation of the Restraint Scale in

ARTICLE IN PRESS
Appetite 49 (2007) 109–121
www.elsevier.com/locate/appet
Research report
Construct validation of the Restraint Scale in normal-weight
and overweight females
Tatjana van Striena,, C. Peter Hermanb, Rutger C.M.E. Engelsc,
Junilla K. Larsena, Jan F.J. van Leeuwed
a
Institute for Gender Studies and Behavioral Science Institute, Radboud University Nijmegen, P.O. Box 9104, 6500 HE Nijmegen, The Netherlands
b
Department of Psychology, University of Toronto, Ont., Canada M5S 3G3
c
Behavioral Science Institute Radboud University Nijmegen, P.O. Box 9104, 6500 HE Nijmegen, The Netherlands
d
Statistical Consultancy Group, Radboud University Nijmegen, P.O. Box 9104, 6500 HE Nijmegen, The Netherlands
Received 13 September 2006; received in revised form 22 December 2006; accepted 2 January 2007
Abstract
The Restraint Scale (RS) is a widely used measure to assess restrained eating. The purpose of this study was to examine the construct
validity of the RS in a sample of normal-weight (n ¼ 349) and overweight (n ¼ 409) females using confirmatory factor analyses of the RS
in relation to other measures for dieting, overeating and body dissatisfaction. Following Laessle et al. [(1989a). A comparison of the
validity of three scales for the assessment of dietary restraint. Journal of Abnormal Psychology, 98, 504–507], we assumed a three-factor
structure: (1) overeating and disinhibitory eating, (2) dieting and restriction of food intake, and (3) body dissatisfaction and drive for
thinness. Analyses revealed that the RS loaded significantly on all three factors for both samples, confirming its multifactorial structure.
However, the RS appears to capture these constructs differently in overweight and normal-weight females such that the RS may
overestimate restraint in overweight individuals. This may explain the greater effectiveness of the RS in predicting counter-regulation in
normal-weight than in overweight samples of dieters.
r 2007 Elsevier Ltd. All rights reserved.
Keywords: Restraint Scale; Dieting; Overeating; Body dissatisfaction; Construct validity
Introduction
A great deal of current research on overeating and eating
disorders has been inspired by restraint theory. This theory
has suggested that dietary restraint or dieting1 (i.e.,
attempted restriction of food intake in order to maintain
or lose weight) contributes to overeating and eating
disorders (Herman, Polivy, & Leone, 2005; Polivy &
Herman, 1985, 1993). This argument is based on various
experiments in which participants scoring high on the
Restraint Scale (RS; Herman, Polivy, Pliner, Threlkeld, &
Corresponding author.
E-mail address: [email protected] (T. van Strien).
Although the term dietary restraint originally referred to a tendency to
oscillate between periods of caloric restriction and overeating (Heatherton, Herman, Polivy, King, & McGree, 1988), we use the term ‘dietary
restraint’ as synonymous with ‘dieting’, avoiding any assumptions about
whether it is associated with overeating.
1
0195-6663/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.appet.2007.01.003
Munic, 1978)—a scale designed to identify dieters—
showed disinhibitive food intake. They displayed elevated
food intake (in a ‘‘taste-test’’ paradigm) when their selfcontrol (inhibition) was undermined by the forced intake of
a forbidden amount or type of food (a preload), alcohol,
distress, and other factors that disrupt self-control
(Herman & Polivy, 2004).
Although disinhibited food intake in restrained eaters is
reasonably well-established in the experimental studies in
which restraint is assessed using the RS,2 a striking contrast
2
Although restrained eaters (as identified by the RS) show disinhibition
as noted above (Herman & Polivy, 2004; Herman et al., 2005), there is
some question about the generality of these effects. For instance, although
a significant restraint-by-preload interaction has been found in seven
classical preload taste-test studies using the RS, the precise pattern of the
interaction does not always support the contention that a forbidden
preload leads restrained eaters to eat significantly more than they would
eat in the absence of the preload. In three of these studies, the interaction
ARTICLE IN PRESS
110
T. van Strien et al. / Appetite 49 (2007) 109–121
appears when restraint is assessed in terms of other
measures—namely, the restraint subscale of the Dutch
Eating Behaviour Questionnaire (DEBQ; Jansen et al.,
1988; Van Strien, Frijters, Bergers, & Defares, 1986a, b;
Wardle & Beales, 1987) or the Three Factor Eating
Questionnaire (TFEQ; Lowe & Kleifield, 1988; Stunkard
& Messick, 1985). Moreover, disinhibited eating by overweight restrained eaters has never been observed in
preload/taste-test studies (Lowe, Foster, Kerzhnerman,
Swain, & Wadden, 2001; McCann, Perri, Nezu, & Lowe,
1992; Ruderman & Christensen, 1983; Ruderman &
Wilson, 1979; Van Strien & Ouwens, 2003a; Wardle &
Beales, 1988), even when the RS is used as the measure of
restraint (McCann et al., 1992; Ruderman & Christensen,
1983; Ruderman & Wilson, 1979).3
This inconsistent pattern has sometimes been attributed
to differences between the three restraint scales. The
suggestion has been made that the DEBQ and the TFEQ
both tend to select a broad range of dieters, successful
dieters and unsuccessful dieters, whereas the RS tends to
select dieters who combine dieting with a tendency to
overeat—unsuccessful dieters4 (Allison, Kalinsky, &
(footnote continued)
arose not so much because preloaded restrained eaters ate significantly
more but because preloaded unrestrained eaters ate significantly less
(Hibscher & Herman, 1977; Jansen, Oosterlaan, Merckelbach, & Van den
Hout, 1988; Ruderman & Christensen, 1983). In one study the interaction
arose because restrained dieters (not quite the same thing as restrained
eaters) ate less following the preload (Lowe, Whitlow, & Bellwoar, 1991).
So RS restrained eaters ate significantly more following the preload in
only three preload studies (Herman & Mack, 1975; Herman, Polivy, &
Esses, 1987; Polivy, Heatherton, & Herman, 1988), although a few other
studies (with additional manipulated variables) have also found evidence
of restrained eaters eating more following a preload than in the absence of
a preload. These varying outcomes do not invalidate the notion that
preloads disinhibit restrained eaters, but they do raise the issue of how
robust the effect is. Indeed, there is reason to expect (Herman & Polivy,
1984) that the disinhibiting effect of a preload will be evanescent,
depending on exactly what preload values are chosen (Herman et al.,
1987). A more reliable means of disinhibiting eating in RS restrained
eaters is to expose them to manipulations of distress based on ‘‘ego
threat’’; i.e., challenges to one’s personal adequacy (Heatherton, Herman,
& Polivy, 1991).
3
The failure of overweight restrained eaters to display counterregulation may reflect a difference in the location of the ‘‘diet boundary’’
(Herman & Polivy, 1984) in overweight and normal-weight restrained
eaters. If, for instance, overweight restrained eaters’ diet boundary were
shifted substantially to the right, toward the satiety boundary—reflecting a
more ‘‘lenient’’ diet, which in turn might be partially responsible for their
overweight—then a given preload would be correspondingly less likely to
violate the diet and induce the disinhibition necessary for counterregulation to emerge (Herman, Polivy, & Van Strien, 2006). It is possible,
in other words, that the failure to find counter-regulation in overweight
restrained eaters in the standard preload paradigm may not reflect a defect
or limitation of the RS, but rather some other aspect of the eating
dynamics of overweight dieters.
4
Different dieting strategies may be associated with success or failure of
restraint. The restraint scales of the TFEQ has been differentiated into two
types of control of eating behavior: flexible (7 items) and rigid (7 items)
control (Westenhoefer, 1991). It may be postulated that ‘‘rigid control’’ is
associated with ‘‘tendency toward overeating.’’ ‘‘Rigid control’’ and
‘‘tendency toward overeating’’ may even be two aspects of the same
Gorman, 1992; Heatherton, et al., 1988; Laessle, Tuschl,
Kotthaus, & Prike, 1989a; Lowe, 1993; Van Strien, 1999;
Van Strien, Breteler, & Ouwens, 2002). The restraint
subscales of the DEBQ and TFEQ measure intended and
actual control/restriction of food intake and have been
shown to have good validity5 with respect to various
measures of food intake (Ard, Desmond, Allison, &
Conway, 2006; De Castro, 1995; Green, Rogers, Elliman,
& Gatenby, 1994; Hill & Robinson, 1991; Laessle et al.,
1989a b; Tuschl, Platte, Laessle, Stichler, & Pirke, 1990;
Van Strien, Frijters, Van Staveren, Defares, & Deurenberg,
1986; Wardle, 1987; Wardle & Beales, 1987; Wardle et al.,
1992). The DEBQ and the TFEQ have specific subscales
for dietary restraint, separate from any tendency to
overeat. By contrast, the RS includes items assessing
restraint and other items assessing disinhibition or overeating, (e.g., ‘‘Do you eat sensibly in front of others and
splurge alone?’’), but these items are not separated into
subscales. The RS has been found to be less clearly
associated with reduced energy intake (French, Jeffery, &
Wing, 1994; Klesges, Isbell, & Klesges, 1992; Klesges,
Klem, & Bene, 1989; Laessle et al., 1989a; Wardle, 1987).
Clear associations have been found, however, between the
RS and measures of binge eating (Hawkins & Clement,
1980; Laessle et al., 1989a; Stice, Ozer, & Kees, 1997;
(footnote continued)
phenomenon. Although some research indicates that the 7-item flexiblecontrol scale predicts a more successful variant of restrained eating,
findings regarding the 7-item control scales have been inconsistent
(Masheb & Grilo, 2002; Shearin, Russ, Hull, Clarkin, & Smith., 1994;
Smith, Williamson, Bray, & Ryan, 1999). Expanded scales of both rigid
and flexible control may be more valid (Timko & Perone, 2005) and
should be examined further. For the current study we did not have
complete data on these expanded scales, so only the 21-item restraint scale
of the TFEQ was used as indicator of more successful restrained eating.
5
Stice, Fischer, and Lowe (2004) questioned the validity of the DEBQ
and TFEQ restraint scale on the basis of outcomes of studies using
unobtrusive measures for food intake, but this conclusion has been refuted
by Van Strien, Engels, van Staveren, and Herman (2006). A major
problem with the study by Stice et al. (2004) is that food intake was
measured only at one moment in time, which is at variance with both the
fundamentals of valid dietary assessment and the concept of restraint as a
trait. In nutritional science, single eating episodes are regarded as
inappropriate for assessing chronic dietary intake: a minimum time
window of 24-h is normally recommended (Stubbs, Johnstone, O’Reilly, &
Poppitt, 1998). Moreover, we know of at least three experimental taste-test
studies where the positive association between restraint and food
consumption disappeared or even became negative when tendency toward
overeating was controlled for (van Strien et al., 2000; Ouwens, Van Strien
& Van der Staak, 2003a, b). In the studies by Stice et al. (2004), the
tendency to overeat was not partialled out of the relation between restraint
and food intake. A separate problem with all validity studies is that
absolute measures of intake may not reflect restriction of food intake, i.e.
eating less than desired. In theory, people may eat more than is required
(in terms of physical activity and body weight) and still be restrained eaters
insofar as (owing to dietary restraint) they eat less than they would
otherwise be inclined to eat. Van Strien et al. (2006) concluded that the
existing restraint scales do in fact validly assess restriction of food intake,
albeit in a more complex fashion than is evident from simple correlations
in single episodes. Stice, Presnell, Lowe, and Burton (2006) have
challenged our conclusions, but further counterargumentation would take
us too far afield.
ARTICLE IN PRESS
T. van Strien et al. / Appetite 49 (2007) 109–121
Wardle, 1980), no doubt because some of its items reflect
overeating while other items reflect undereating. Indeed,
some critics (Charnock, 1989; Stice et al., 1997; Wardle &
Beales, 1987) have castigated the RS for ‘‘criterion
confounding,’’ arguing that people who score high on the
RS are more likely to overeat in laboratory studies because
the RS itself contains items that tap the propensity to
overeat.6
Support for the supposition that the restraint scales may
identify a different sort of dieter than do the other restraint
scales was found in Laessle et al. (1989a)’s factor analysis
of the RS in relation to other measures of dieting,
overeating and body dissatisfaction. Laessle et al. (1989a)
found that the three restraint scales measured different
constructs, and that neither the DEBQ-R nor the TFEQ-R
shared a common factor with the RS. Furthermore, the RS
was found to be closely related to disinhibited eating and
weight fluctuation, but not to successful caloric restriction.
In contrast, the DEBQ-R and TFEQ-R represented more
successful dieting behavior. Still, the Laessle et al. study
was limited: the analyses were based on an exploratory
factor analysis of only 60 normal-weight college women.
Stable factor structures can be obtained only with samples
of over 300 subjects (Tabachnick & Fidell, 2001).
Factor analyses of the RS on the level of its items
typically reveal two factors—Concern for Dieting (CD)
and Weight Fluctuation (WF) (Allison et al., 1992; Van
Strien et al., 2002). In overweight samples, however,
additional factors tend to emerge; perhaps the RS measures
a different set of constructs in overweight individuals
(Ruderman, 1986). Further, some investigators have
argued that WF items may distort RS scores in the
overweight. Because of variability in activity level and/or
spontaneous diuresis, overweight people’s weight tends to
fluctuate more than does that of normal-weight individuals
(Field et al., 2004). Because the WF items of the RS are
scored in terms of absolute (not proportional) weight
changes, overweight individuals may obtain high scores on
the RS simply on the basis of large weight fluctuations,
even if they do not watch their weight or consciously
restrain their eating (Rodin, 1981).7 The RS has been
6
According to Heatherton et al. (1988), what some call ‘‘criterion
confounding’’ others might with justification call ‘‘construct validity.’’
Herman and Polivy intended the RS to assess overeating as well as
undereating, because they started from the assumption that dieters
alternate between overeating and undereating. Although in principle
dieting should be strictly matter of under eating, in reality most dieters’
under eating is punctuated by episodes of overeating.
7
Heatherton, Polivy, and Herman (1991) found that weight fluctuates as
a function of RS scores more than as a function of weight per se. Is this
finding a reflection of ‘‘criterion confounding’’ (i.e., the RS includes items
assessing weight fluctuation, so that restrained eaters will necessarily
display more weight fluctuation because of the content of the RS)?
Heatherton et al. found that the CD subfactor of the RS (which does not
include weight-fluctuation items) was as strongly associated with their
measure of maximum weight fluctuation as was the WF (Weight
Fluctuation) subscale. Given that restrained eaters display higher weight
fluctuations without the confounding influence of weight-fluctuation
items, it seems likely that the association between restraint and weight
111
repeatedly used in different samples (e.g., normal-weight
and overweight individuals (Lowe, 1993; Ruderman &
Christensen, 1983). An adequate examination of the
construct validity in different samples is therefore warranted—indeed, overdue (Lowe, 2002). The main purpose
of the present study was to examine the construct validity
of the RS in relation to other measures of dieting and
overeating in large samples of normal-weight and overweight females, using confirmatory factor analysis. Confirmatory methodology has unique advantages over
exploratory analyses because measurement models are
developed on an a priori basis and specific factor structures
can be tested to see whether they fit the data.
Following Laessle et al. (1989a) we assume for both the
overweight and normal-weight females that the three
restraint scales measure different constructs, and that the
RS shows a three factor structure: (1) overeating and
disinhibitory eating (2) dieting and restriction of food
intake and (3) body dissatisfaction and drive for thinness.
Further it is an open empirical question whether the RS
captures these constructs differently in overweight compared to normal-weight females.
Method
Participants
From a larger sample of female university students
(n ¼ 411), 349 normal-weight women (p18 BMI p25)
(BMI ¼ Body Mass Index (weight/heightheight) (body
weight and height self-reported) for whom we had
complete data were selected to participate in Study 1. We
excluded students with clear overweight (BMIX26) in
order to optimize comparability with previous experimental studies on dietary restraint that have been predominantly concentrated on normal weight university samples.
They had been recruited at the campus of the Radboud
University Nijmegen, most of them were undergraduates,
and most of them studied psychology, medicine, literature
or law. Mean BMI in the Study 1 sample was 21.3
(SD ¼ 1.8) and mean age was 20.9 years (SD ¼ 2.2). From
a larger sample (n ¼ 824), 409 overweight women
(26pBMI p40; agep60 years) for whom we had complete
data were selected to participate in Study 2. These women
were recruited through advertisements (a call for obese
persons who were offered a personal eating diagnosis in
return for their participation) in local newspapers and
obesity bulletins. We excluded females with morbid obesity
(BMI440) in order to optimize comparability with female
(footnote continued)
fluctuation arises not from a confound but rather from restrained eaters’
alternating periods of weight loss and weight regain. By the same token,
overweight people probably score higher on the RS not because of the
‘‘confounding’’ influence of weight fluctuation items, but because (a)
overweight people are more invested in weight loss (dietary restraint) than
are normal-weight people (Lowe, 1984) and/or (b) people who are invested
in dietary restraint and weight loss are more likely to gain weight?
ARTICLE IN PRESS
T. van Strien et al. / Appetite 49 (2007) 109–121
112
participants in weight loss treatments (morbid obese are
offered different treatment (surgery) in the Netherlands).
Mean BMI was 31.9 (SD ¼ 3.8), mean age was 40.2 years
(SD ¼ 11.3), 7.4% had only primary education, 44.7% had
secondary education, and 47.2% had tertiary education.
The overweight sample was clearly older and also had a
lower level of education than did the normal-weight
sample. Participants completed various Dutch versions of
measures of dieting, overeating, and body dissatisfaction.
Measures
1. The RS (Herman et al., 1978; Dutch version: Jansen
et al., 1988) assesses dieting and weight fluctuation (10
items; Table 1).
In earlier studies of the factor structure of the RS on the
level of its items (Blanchard & Frost, 1983; Lowe, 1984;
Overduin & Jansen, 1996) two factors (CD: concern for
dieting and WF: weight fluctuation) were identified (see
Table 1). Van Strien et al. (2002) also identified these
two subscales, but two items were not included in the
van Strien et al. (2002) subscales: one item (item 6) was
dropped because it clearly refers to disinhibited eating
(Wardle & Beales, 1987), an alleged instance of criterion
confounding (Stice et al., 1997), and one item (item 10)
was dropped because there had been little consensus in
earlier studies as to its proper factor assignment
(Blanchard & Frost, 1983; Lowe, 1984; Overduin &
Jansen, 1996). In these earlier studies, item 10 sometimes
loaded on the CD factor and sometimes on both the CD
and WF factors. A further feature of item 10 is that it
Table 1
Subscale structure in the restraint scale (RS)
CD
WF
WF
WF
CD
CD
CD
CD
CD
WF
1. How often are you dieting?
Never; rarely, sometimes, often, always (Scored 0–4)
2. What is the maximum amount of weight (in kilos) you
have ever lost within 1 month?
(0–2.5; 2.5–5; 5–7.5; 7.5–10; 10+ (Scored 0–4))
3. What is the maximum amount of weight gain (in kilos)
within a week?
(0–0.5; 0.5–1; 1–1.5; 1.5–2.5; 2.5+ (Scored 0–4))
4. In a typical week, how much does your weight fluctuate?
(0–0.5; 0.5–1; 1–1.5; 1.5–2.5; 2.5+ (Scored 0–4))
5. Would a weight fluctuation of 2.5 kilos affect the way you
live your life?
Not at all; slightly, moderately; very much (Scored 0–3)
6. Do you eat sensibly in front of others and splurge alone?
Never; rarely, often, always (Scored 0–3)
7. Do you give too much time and thought to food?
Never, rarely, often; always (Scored 0–3).
8. Do you have feelings of guilt after overeating?
Never, rarely, often, always (Scored 0–3).
9. How conscious are you what you are eating?
Not at all; slightly, moderately, extremely (Scored 0–3)
10. How many kilos over your desired weight were you at
your maximum weight?
(0–0.5; 0.5–3; 3–5; 5–10; 10+ (Scored 0–4).
Note: CD ¼ concern for dieting; WF ¼ weight fluctuation.
refers to a history of overweight, which can be
considered different from weight fluctuation (Lowe,
1984).
These subscales have been shown to display factorial
unidimensionality and adequate internal consistency in
a sample of normal-weight females (van Strien et al.,
2002). Exploratory factor analysis for the present
normal-weight sample yielded two factors with the five
CD items 1, 5, 7, 8, and 9 and three WF items 2, 3, and
4. Confirmatory factor analysis supported this result
both for the normal-weight and the overweight sample 8.
For the normal-weight sample, Cronbach’s alpha was
0.81 for the 5-item CD scale (all CD items listed in
Table 1 except for item 6), 0.68 for the 3-item WF scale
(all WF items listed in Table 1 except for item 10), and
0.84 for the total RS10 scale. The Cronbach’s alpha
coefficients for the overweight sample were 0.65, 0.72,
and 0.73 for the 5-item CD scale, the 3-item WF scale,
and the RS10 scale, respectively.
2. The Dutch Eating Behavior Questionnaire (DEBQ;
original Dutch version: Van Strien et al., 1986a, b;
English version: Van Strien, 2002a) has 33 items,
forming three separate scales: emotional eating (13
items; e.g. ‘‘Do you have a desire to eat when you are
irritated?’’), external eating (10 items; e.g. ‘‘If food
smells and looks good, do you eat more than usual?’’),
and restrained eating (10 items; e.g. ‘‘Do you try to eat
less at mealtimes than you would like to eat?’’).
Response categories range from 1 (‘never’) to 5 (‘very
often’). Each of the scales displayed good internal
consistency and factorial validity (e.g., Van Strien, 1996;
Van Strien et al., 1986a, b), high convergent and
discriminant validity (Van Strien, 2002a) and good
validity for food consumption (Van Strien, 2005). For
the normal-weight sample, Cronbach’s alpha was 0.89
for emotional eating, 0.74 for external eating, and 0.93
for restrained eating. The reliabilities (Cronbach’s
alphas) for the overweight sample were 0.94, 0.84, and
0.89 for emotional eating, external eating, and restrained
eating, respectively.
3. The Three Factor Eating Questionnaire (TFEQ;
Stunkard & Messick, 1985; Dutch version: Van Strien,
Cleven, & Schippers, 2000) has 51 items forming three
separate scales: cognitive restraint (21 items; e.g. ‘‘I
deliberately take small helpings as a means of controlling my weight’’), disinhibition (16 items; e.g. ‘‘While on
a diet, if I eat a food that is not allowed, I often then
splurge and eat other high calorie foods’’), and hunger
(14 items; e.g. ‘‘I am usually so hungry that I eat more
than three times a day’’). Thirty-six items have a truefalse response format; the other 15 items have varying
8
For the normal-weight sample, chi-square ¼ 42.31, df ¼ 19, po.003,
GFI ¼ .97, AGFI ¼ .94, NFI ¼ .95, RMSEA ¼ .06. For the overweight
sample, chi-square ¼ 55.96, df ¼ 19, po.001, GFI ¼ .97, AGFI ¼ .94,
NFI ¼ .91, RMSEA ¼ .07 (see the Analysis and Results section for the
interpretation of the fit indicators).
ARTICLE IN PRESS
T. van Strien et al. / Appetite 49 (2007) 109–121
response options (e.g., rarely [1] to always [4], or easy [1]
to difficult [4]). Exploratory factor analyses have not
always replicated the original three-factor structure
(Ganley, 1988; Hyland, Irvine, Thacker, Dann, &
Dennis, 1989). Confirmatory methods have also shown
poor replication of the proposed factor structure
(Mazzeo, Aggen, Anderson, Tozzi, & Bulik, 2003).
For the normal-weight sample, Cronbach’s alpha was
0.88 for cognitive restraint, 0.72 for disinhibition, and
0.69 for hunger. The reliabilities for the overweight
sample were 0.67, 0.63, and 0.66 for cognitive restraint,
disinhibition, and hunger, respectively.
4. The Eating Disorder Inventory Revised (EDI-II; Garner, 1991; Dutch version: Van Strien, 2002b) has 91
items forming 11 scales. For the present study, in line
with the study by Laessle et al. (1989a), only three scales
were used: bulimic eating (7 items; e.g., ‘‘I think about
bingeing’’), body dissatisfaction (9 items; e.g., ‘‘I think
that my stomach is too big’’), and drive for thinness (7
items; e.g., ‘‘I am terrified of gaining weight’’). The body
dissatisfaction and drive-for-thinness scales were included as measures of motivational variables thought to
lead to dietary restraint. They refer to issues directly
related to weight control, such as preoccupation with
weight and unhappiness with one’s thighs and hips.
Response categories range from 1 ‘never’ to 6 ‘always.’
In contrast with the EDI manual (Garner, 1991), in
which a transformation of responses into a four-point
scale is advocated, the present study utilized untransformed responses, because scale transformation was
found to damage the validity of the EDI among a nonclinical population (Schoemaker, Van Strien, & Van der
Staak, 1994; Van Strien & Ouwens, 2003b). For the
normal-weight sample, Cronbach’s alpha was 0.85 for
bulimic eating, 0.95 for body dissatisfaction, and 0.93
113
for drive for thinness. The reliabilities for the overweight
sample were 0.89, 0.83, and 0.89 for bulimic eating,
body dissatisfaction, and drive for thinness, respectively.
In addition, the following two questions were administered: ‘‘Are you currently dieting?’’ and ‘‘Have you ever
had an eating binge, i.e., you ate an amount of food others
would consider unusually large?’’ Both questions had
dichotomous answer categories (0 ¼ no, 1 ¼ yes) (see also
Lowe (1993), Lowe and Timko ( 2004), Van Strien, Engels,
Van Leeuwe and Snoek ( 2005)).
Analyses and results
Differences in means between normal-weight and overweight
females
In Table 2 means and standard deviations of all scales
are presented. Differences between the two groups were
tested by t-tests not assuming equal variances; effect sizes
are also included. Means of overweight females are higher
on all scales except for the external eating scale of the
DEBQ.
Correlations between RS and measures of dieting,
overeating, and body dissatisfaction
Table 3 shows the Pearson correlation coefficients of all
variables for the sample of normal-weight and overweight
females. Differences between normal and overweight
correlations were tested by two-sided r-to-z tests, Bonferroni corrected, and significant differences in the correlations between the two groups can be found in the second
part of Table 3. Of special interest for the present study are
the correlations between the RS (RS10, and the CD and
Table 2
Means, standard deviations and t-test for the two samples
1
2
3
4
5
6
7
8
9
10
11
12
13
Scale
Normal weight
N ¼ 349
Mean
DEBQ-em
DEBQ-ex
TFEQ-dis
EDI-Bu
Ever binge?b
DEBQ-R
TFEQ-R
Do you diet?b
EDI-DT
EDI-BD
RS10
RS-CD
RS-WF
2.64
3.14
1.46
11.34
0.12
2.61
1.68
0.21
17.14
31.28
1.91
1.88
2.05
SD
Overweight
N ¼ 409
Mean
Difference
SD
T
p
da
0.58
0.47
0.23
4.27
0.33
0.83
0.36
0.41
7.54
9.77
0.53
0.53
0.63
3.13
3.08
1.65
16.32
0.37
3.19
2.04
0.54
27.4
46.75
2.86
2.74
2.70
0.83
0.61
0.15
7.48
0.48
0.69
0.21
0.51
7.01
7.34
0.51
0.55
0.58
9.7
1.4
13.1
11.4
8.4
10.6
16.3
10.1
19.3
24.3
25.1
21.3
14.6
o0.001
0.157
o.0.001
o0.001
o0.001
o0.001
o0.001
o0.001
o0.001
o0.001
o0.001
o0.001
o0.001
0.68
+0.11
1.00
0.80
0.60
0.78
1.26
0.72
1.41
1.81
1.85
1.59
1.08
em ¼ emotional eating; ex ¼ external eating; dis ¼ disinhibition; Bu ¼ bulimia; R ¼ restraint; DT ¼ drive for thinness; BD ¼ body dissatisfaction;
CD ¼ concern for dieting; WF ¼ weight fluctuation.
a
d ¼ effect size by Cohen’s d (0.20 ¼ small, 0.50 ¼ medium, 0.80 ¼ large).
b
0 ¼ no, 1 ¼ yes.
ARTICLE IN PRESS
T. van Strien et al. / Appetite 49 (2007) 109–121
114
Table 3
Intercorrelations between scales for normal weight and overweight females
Scale
1
2
3
4
Normal weight females (N ¼ 349)b
5
6
7
8
9
10
11
12
13
1
2
3
4
5
6
7
8
9
10
11
12
13
DEBQ-em
DEBQ-ex
TFEQ-dis
EDI- Bu
Ever binge?
DEBQ-R
TFEQ- R
Do you diet?
EDI-DT
EDI-BD
RS10
RS-CD
RS-WF
—
0.27
0.22
0.40
0.60
—
0.15
0.03
0.42
0.36
0.14
—
0.11
0.05
0.41
0.37
0.12
0.86
—
0.00
0.08
0.20
0.14
0.05
0.46
0.49
—
0.25
0.15
0.50
0.59
0.27
0.75
0.79
0.45
—
0.24
0.09
0.41
0.45
0.20
0.51
0.51
0.25
0.67
—
0.28
0.13
0.59
0.65
0.35
0.71
0.74
0.44
0.82
0.64
—
0.25
0.10
0.52
0.57
0.28
0.77
0.82
0.48
0.89
0.60
0.89
—
0.15
0.08
0.37
0.42
0.27
0.31
0.32
0.24
0.36
0.35
0.72
0.39
—
1
2
3
4
5
6
7
8
9
10
11
12
13
DEBQ-em
DEBQ-ex
TFEQ-dis
EDI- Bu
Ever binge?
DEBQ-R
TFEQ- R
Do you diet?
EDI-DT
EDI-BD
RS10
RS-CD
RS-WF
0.41
0.30
0.19
0.68
—
0.03
0.11
0.06a
0.00a
0.01
—
0.04
0.12
0.07a
0.03a
0.04
0.66a
—
0.06
0.02
0.00
0.04
0.04
0.42
0.38
—
0.29
0.16
0.11a
0.44
0.29
0.44a
0.39a
0.28
—
0.18
0.20
0.09a
0.22a
0.18
0.04a
0.13a
0.11
0.41a
—
0.42
0.34
0.16a
0.61
0.42
0.36a
0.35a
0.30
0.62a
0.32a
—
0.32
0.25
0.09a
0.45
0.28
0.55a
0.53a
0.40
0.75a
0.31a
0.81a
—
0.25
0.20
0.10a
0.37
0.27
0.04a
0.07a
0.07
0.21
0.17
0.76
0.27
—
0.42
—
0.56
0.39
—
0.51
0.31
0.69
—
Overweight Females (N ¼ 409)c
—
0.48
0.16a
0.67
0.48
—
0.03a
—
0.28a
—
a
Difference between Normal and Overweight correlation is significant (5%-level, two-sided r-to-z test, Bonferroni corrected).
If N ¼ 349 a correlation (r) is significant at the 5% level if r40.105 and at the 1% level if r40.137.
c
If N ¼ 409 a correlation (r) is significant at the 5% level if r40.097 and at the 1% level if r40.126.
b
WF subscales) and measures of overeating (DEBQ emotional eating, DEBQ external eating, TFEQ disinhibited
eating, EDI bulimia and the question ‘‘have you ever had
an eating binge’’), dieting (DEBQ restrained eating, TFEQ
restrained eating, and the ‘‘Do you diet?’’ question), and
body dissatisfaction (EDI drive for thinness and body
dissatisfaction) in the different samples.
Study 1: Normal-weight sample. In the sample of normalweight female students, RS10 showed high correlations
(r40.50) with two of the measures of overeating (TFEQ
disinhibition and EDI-bulimia). Correlations with two
measures of restrained eating (DEBQ restraint and TFEQ
restraint) were even higher (r40.70), but the highest
correlation of the RS10 (r40. 80) was obtained with one
of the indicators of body dissatisfaction (EDI-drive for
thinness). Highly similar patterns of results were obtained
with the CD subscale. The WF subscale showed its highest
correlations with measures of overeating (EDI-bulimia and
TFEQ disinhibition).
Study 2: Overweight sample. In the sample of overweight
females, RS10 showed its highest correlations with a
measure of overeating (EDI-bulimia) and a measure of
body dissatisfaction (EDI-drive for thinness; rX0.60;
Table 2). Correlations with the measures of restrained
eating (DEBQ restraint and TFEQ restraint) were lower
(ro0.40). In the overweight sample the subscale CD
showed its highest correlations with measures of restrained
eating (DEBQ restraint and TFEQ restraint; r40.50), and
the subscale WF correlated most strongly with measures of
overeating (EDI-bulimia and the ‘‘Ever binge’’ question;
r40.25) and a measure of body dissatisfaction (EDI-drive
for thinness; r ¼ 0.22).
Confirmatory factor analysis
Confirmatory factor analysis was performed by AMOS
5.0 (Arbuckle & Wothke, 1999). A three-factor structure
was assumed to exist, as in Laessle et al.’s (1989a) study: (1)
overeating and disinhibitory eating, (2) dieting and
restriction of food intake, and (3) body dissatisfaction
and drive for thinness. The five overeating measures
(DEBQ Emotional Eating, DEBQ External Eating, TFEQ
Disinhibition, EDI Bulimia, and the ‘‘Ever binge?’’
question) were assumed to load on the first factor
(overeating). The three dieting measures (DEBQ Restraint,
TFEQ Restraint and the ‘‘Currently dieting?’’ question)
were assumed to load on the second factor (dieting), and
EDI drive for thinness and body dissatisfaction were
ARTICLE IN PRESS
T. van Strien et al. / Appetite 49 (2007) 109–121
supposed to comprise the third factor (body dissatisfaction). The result of the confirmatory factor analysis with
RS10 loading on each of the three factors was contrasted to
the result of the analysis with RS10 loading on only the
second factor. Evidence that RS10 is loading on each of the
three factors (the multifactorial hypothesis; for a diagram
of the model to be tested, see Fig. 1) can be provided by the
fit of the model with the three factor loadings and the
significance of the chi-square difference between the two
models. This procedure is repeated for the two subscales
CD and WF.
This procedure was applied to a sample of normalweight females (Study 1) and a sample of overweight
females (Study 2). Fit of the factor model was judged by
using the chi-square test, the goodness-of-fit index (GFI),
the adjusted goodness-of-fit index (AGFI), the normed fit
index (NFI) and the root mean square error of approximation (RMSEA). A model fits reasonably well if the chi-
115
square value does not exceed a limited multiple of its
degrees of freedom, if the GFI, AGFI, and NFI are greater
than 0.90, and if the RMSEA is smaller than 0.08 (Hu &
Bentler, 1999).
Study 1: Normal-weight sample. In an initial CFA, the
scales 1–10 of Table 2 were included: five of them were
supposed to load on overeating, three on dieting and two
on body dissatisfaction according to the left part of the
diagram in Fig. 1. In the test of the initial model, the
disinhibition scale of the TFEQ was responsible for a
substantial lack of fit. Fit improved considerably when this
variable was removed. Additional improvement of the fit
could be achieved by freeing the covariance between the
error terms of DEBQ emotional eating and DEBQ external
eating. These error terms might be correlated owing to
shared origin and format. The fit of this 9-scale model
without RS10 was satisfactory: w2[25] ¼ 50.76 (see also the
first row of Table 4). To check whether the restraint scale
TFEQ-dis
DEBQ-em
DEBQ-ex
Overeating
EDI-Bu
Ever binge ?
DEBQ-R
TFEQ-R
RS10
Dieting
Do you diet ?
EDI-DT
Body
Dissatisfaction
EDI-BD
Fig. 1. A diagram of the model to be tested.
Table 4
Model fit for confirmatory factor analyses
Sample and model
w2
Normal-weight sample (n ¼ 349)
9-scale model
50.76
RS10
78.54
CD&WF
73.26
Overweight sample (n ¼ 409)
10-scale model
62.06
RS10
81.08
CD&WF
97.39
df
p
GFI
AGFI
NFI
RMSEA
25
31
38
o0.002
o0.001
o0.001
0.97
0.96
0.96
0.94
0.93
0.94
0.97
0.96
0.97
0.05
0.07
0.05
33
40
48
o0.002
o0.001
o0.001
0.97
0.96
0.96
0.95
0.94
0.93
0.95
0.95
0.95
0.05
0.05
0.05
Note: GFI ¼ goodness of fit index; AGFI ¼ adjusted goodness of fit index; NFI ¼ normed fit index; RMSEA ¼ root mean square error of
approximation.
ARTICLE IN PRESS
T. van Strien et al. / Appetite 49 (2007) 109–121
116
of the DEBQ measures overeating and body dissatisfaction
too, the corresponding loadings were added accordingly.
Estimates of these loadings were very small and not
significant (0.04 for overeating and 0.07 for body
dissatisfaction, respectively). Moreover, the chi-square
hardly diminished at all (w2[23] ¼ 49.93) and the chi-square
difference test was not significant (w2[2] ¼ 0.83). The same
procedure was applied to the restraint scale of the TFEQ.
Estimates of the loadings were small and not significant
(0.02 for overeating and 0.15 for body dissatisfaction,
respectively). The chi-square hardly diminished at all
(w2[23] ¼ 50.00) and the chi-square difference test was not
significant (w2[2] ¼ 0.76). Hence the restraint scales of both
the DEBQ and the TFEQ appear to measure dieting only.
Next, the same procedure was applied to the RS10 scale. A
model with RS10 loading on dieting only yielded
w2[33] ¼ 236.45, whereas the model with RS10 loading on
overeating, dieting, and body dissatisfaction (as depicted in
Fig. 1) gave: w2[31] ¼ 78.54 (see also Table 4, RS10 model).
Thus, the chi-square difference test was highly significant,
confirming the indispensability of the two loadings of RS10
on overeating and body dissatisfaction.
It should be stressed that the correlations between the
three factors were all significant (po0.01 level) and quite
high: 0.39 for the correlation between overeating and
dieting, 0.83 between dieting and body dissatisfaction, and
0.59 between overeating and body dissatisfaction.
We also estimated the model for the RS subscales CD
and WF instead of RS10. The correlations between the
three factors (0.39, 0.83, and 0.59, respectively) were quite
similar to those estimated in the RS10 model (see Table 5).
The fit for this CD&WF model was satisfactory (see Table
4, third row). Table 5 indicates that the three-factor
multifactorial hypothesis holds perfectly for the CD
subscale but not for the WF subscale (no significant
loading on the body-dissatisfaction factor).
Study 2: Overweight sample. The same analyses were
performed on the data of the overweight females. The
initial model did not fit the data of the overweight subjects
particularly well: w2[34] ¼ 89.02. Again, the fit could be
improved by freeing the covariance between the error terms
of DEBQ emotional eating and DEBQ external eating.
Then w2[33] ¼ 62.06. The total set of fit indices is presented in
Table 4 (10-scale model; overweight females).
Since the overweight sample was older than was the
normal-weight sample, it is worth examining whether the
difference between the overweight and normal-weight
sample might be due to age rather than to degree of
overweight. Accordingly, we split the overweight sample
into older and younger halves. For both the younger
Table 5
Factor loadings for confirmatory factor models with the total restraint scale (RS10) and RS subscales (RSCD and RSWF)
Variables/factors
Normal-weight females
DEBQ-em
DEBQ-ex
EDI-Bu
Ever binge?
DEBQ-R
TFEQ-R
Do you diet?
EDI-DT
EDI- BD
RS10
RSCD
RSWF
Overweight females
DEBQ-em
DEBQ-ex
TFEQ-Dis
EDI-Bu
Ever binge?
DEBQ-R
TFEQ-R
Do you diet?
EDI-DT
EDI-BD
RS10
RSCD
RSWF
RS10
Overeating
Dieting
Body Dis.
0.51a
0.31a
1.00a
0.60a
RSCD/RSWF
Overeating
0.33a
0.90a
0.95a
0.53a
0.97a
0.70a
0.40a
0.99a
0.68a
0.13a
0.30a
0.67a
0.48a
0.28a
1.00a
0.68a
Note: Body Dis. ¼ Body Dissatisfaction.
a
po0.01x.
0.40a
0.28a
0.48a
0.09
0.66a
0.48a
0.27a
1.00a
0.67a
0.83a
0.79a
0.51a
0.51a
Body Dis.
0.51a
0.31a
1.00a
0.60a
0.91a
0.95a
0.52a
0.28a
Dieting
0.34a
0.80a
0.79a
0.53a
1.00a
0.41a
0.23a
1.00a
0.42a
0.30a
0.35a
0.51a
0.12
0.35a
0.00
ARTICLE IN PRESS
T. van Strien et al. / Appetite 49 (2007) 109–121
females (less or equal to 40 years old, N ¼ 195,
w2[33] ¼ 45.56, po0.08, RMSEA ¼ 0.044) and the older
females (older than 40, N ¼ 214, w2[33] ¼ 60.06, po0.004,
RMSEA ¼ 0.062) the 10-scale model fitted well. By
comparing the multigroup solution without restrictions to
the results obtained by restricting the (unstandardized)
loadings to be equal, the chi-square difference test yielded:
w2[10] ¼ 16.38, po0.09, indicating no difference between the
factor structure for younger and older overweight females.
We may conclude that the age difference is unlikely to
explain the normal-weight/overweight difference. Thus we
may analyze the overweight sample as a whole.
The model with DEBQ-R loading on each factor yielded
w2[31] ¼ 61.84, with a non-significant difference in chi-square
and non-significant loadings: 0.02 for overeating and 0.00
for body dissatisfaction. The model with TFEQ-R loading
on each factor yielded w2[31] ¼ 60.93, with a non-significant
difference in chi-square and non-significant loadings:
0.03 for overeating and 0.02 for body dissatisfaction.
Thus, once again, the restraint scales of both the DEBQ
and the TFEQ may be considered to measure dieting only.
Next, the same procedure was applied to the RS10 scale.
The model with RS10 loading on dieting only yielded:
w2[42] ¼ 347.72, whereas the model with RS10 loading on
overeating, dieting, and body dissatisfaction gave
w2[40] ¼ 81.08 (see also Table 4, RS10 model). The chisquare difference test was highly significant, showing the
indispensability of the two loadings of RS10 on overeating
and body dissatisfaction.
The correlation between overeating and dieting was
quite low (0.01) and not significant. The other correlations were significant: 0.52 between dieting and body
dissatisfaction, and 0.44 between overeating and body
dissatisfaction, respectively.
We also estimated the model for the RS subscales CD
and WF instead of RS10. The correlations between the
three factors (0.01, 0.51 and 0.45, respectively) were quite
similar to those estimated in the RS10 model (see Table 5).
The fit for this CD&WF model was satisfactory (see
Table 4). Table 5 indicates that the three-factor multifactorial hypothesis holds perfectly for the CD subscale but
not for the WF subscale (no significant loadings on the
body-dissatisfaction factor and the dieting factor).
The solutions for the two samples show some differences. Considering only the RS10 analysis (first columns in
Table 5), the loadings on the overeating factor are higher
for the overweight sample compared to the normal-weight
sample. The opposite tendency is true for the dieting factor.
For the body-dissatisfaction factor it is worthy of note that
EDI-BD contributes more in the normal-weight sample. If
we leave out TFEQ-Dis we may compare the two solutions.
The multigroup analysis without restrictions (a simultaneous confirmatory factor analysis) yielded a satisfactory
fit (w2[60] ¼ 131.94, po0.001, GFI ¼ 0.967, AGFI ¼ 0.939,
NFI ¼ 0.966, RMSEA ¼ 0.040), indicating that the factor
structure is similar in both groups. The same variables
loaded on the same factors. If we restrict the (unstandar-
117
dized) loadings so that they are equal in the two groups, we
get w2[72] ¼ 433.37, and the chi-square difference test yields
w2[72] ¼ 301.42, which is highly significant. This means that
the magnitude of the loadings cannot be considered equal
across both groups. So we may conclude that the factorial
structure of restraint variables is the same where the
magnitude of the loadings differs. Differences at the scale
level between the correlations in both groups can be found
in Table 3 (second part).
Discussion
The present study confirmed the multifactorial structure
of the RS. In both the sample of normal-weight female
college students and the sample of overweight females, the
RS loaded significantly on the same three factors, viz.,
dieting, overeating, and body dissatisfaction, which is in
line with the results of Laessle et al.’s (1989a) exploratory
factor-analytic study in normal-weight females. As the RS
was found to be closely related to dieting, as well as to
disinhibited eating and body dissatisfaction, we may
suggest that the RS tends to select dieters who exhibit
disinhibited eating (i.e, unsuccessful dieters). We may
conclude that the inconsistent support for the restraint/
disinhibition effect in the experimental research literature is
probably due to differences between the various restraint
scales in the type of dieter that they select (Allison et al.,
1992; Heatherton et al., 1988; Laessle et al., 1989a; Lowe,
1993; Van Strien, 1999; Van Strien et al., 2002). In contrast
to the multidimensional RS, the restraint scales of the
DEBQ and TFEQ seem to be one-dimensional, as they
both showed a satisfactory fit when loading on the Dieting
factor only in the current study.
The RS Concern-for-Dieting subscale had likewise a
multifactorial structure in both samples. The significant
loading of the CD subscale on the Overeating factor is
remarkable in view of the fact that the item that clearly
referred to disinhibited eating (Item 6; see Table 1) was not
included in the 5-item CD measure. Even without this item,
CD show significant associations with overeating.
In the sample of normal-weight female students, the
most pronounced loading of the RS was on body
dissatisfaction. In contrast, in the sample of overweight
females, the RS loaded highest on overeating, possibly due
to the RS’s weight-fluctuation items. In this sample, weight
fluctuation was associated only with overeating. This
finding echoes the concern raised by Rodin (1981) that
the RS weight-fluctuation items may be problematic for the
assessment of restraint in overweight individuals because
some overweight individuals may accrue a high score on
the RS simply on the basis of large weight fluctuations,
rather than because they are currently watching their
weight or consciously restraining their eating. Alternatively, it may be that the RS is not ‘‘contaminated’’ for
overweight individuals but that those restrained eaters who
are most inclined to overeat are the ones who are most
likely to become overweight.
ARTICLE IN PRESS
118
T. van Strien et al. / Appetite 49 (2007) 109–121
The fact that women in the overweight sample were
generally older than were women in the normal-weight
sample raises the possibility that the psychometric differences that we found in the two samples might be due to age
(or history) rather than to degree of overweight per se. Our
subdividing the overweight sample into older and younger
subsamples and finding that these two subsamples did not
differ significantly with respect to the examined psychometrics of their RS scores supports the conclusion that
whatever differences exist between the normal-weight and
the overweight samples are not due to differences in age (or
history) per se.
Although the association between dieting and overeating
was absent in the sample of overweight females, it was
present in the sample of normal-weight female students.
This pattern corresponds to the absence (in overweight
samples) and presence (in normal-weight samples) of a
counter-regulation effect in taste-test experiments. Even
when the RS has been the operative measure of restraint,
no counterregulation effect has appeared in studies using
overweight samples (McCann et al., 1992; Ruderman &
Christensen, 1983; Ruderman & Wilson, 1979).
Our results suggest that we may adopt a different
perspective on the overweight. Insofar as the overweight
are chronic overeaters, it may be that this overeating is not
simply a matter of the disinhibition of prior restraints
(since overeating is not associated with dieting among the
overweight). This line of thought does not invalidate the
notion that restraint may lead to disinhibition (overeating),
but challenges the notion that all overeating requires prior
restraint. This proposal corresponds to the suggestion
(Herman & Polivy, in press; Herman et al., 2005) that there
are ‘‘two routes to overeating,’’ only one of which involves
disinhibition of restrained eating. 9
Herman and Mack (1975), following Nisbett (1972),
suggested that differences in eating patterns between obese
and normal-weight individuals might reflect the fact that
obese individuals are more likely to be restrained eaters.
The pattern of intake that Herman and Mack discovered in
restrained eaters, however, did not correspond to what they
had expected on the basis of prior work on the obese,
which had shown that obese individuals are unresponsive
to a preload manipulation, whereas normal-weight individuals respond in a normal regulatory fashion, i.e., eating
less after a large preload than after a small preload or no
preload (Schachter, Goldman, & Gordon, 1968). Herman
and Mack’s restrained eaters, far from being unresponsive
9
The second route to overeating involves what Herman and colleagues
refer to as ‘‘elevated intake norms’ that induce excessive eating in almost
everyone. In addition there also may be other, more pathological routes to
overeating and weight gain, such as emotional overeating which is in turn
highly associated with binge eating (Van Strien, et al., 2005; Van Strien &
Ouwens, 2007). In these other routes, dieting may be a proxy for other
important predictors of eating disorders and weight gain (see also: Hill,
2004; Lowe & Levine, 2005; Johnson & Wardle, 2005, de Lauzon-Gaulain,
Basdevant, Romon, Karlsson, Borys, Charles and the FLVS Study
Group, 2006; Presnell & Stice, 2003).
to preload size (as demanded by the restraint-obesity
parallel) responded by ‘‘counter-regulating’’ (i.e., eating
more after a rich preload than after no preload). Ironically,
then, the notion that the restrained eaters ought to behave
like obese eaters was not supported, but the quest to
document the parallel led to the discovery of the counterregulatory pattern that has subsequently formed the basis
of studies of restraint and disinhibition in normal-weight
individuals. The pattern that Schachter, Nisbett, and
others ascribed to the obese (i.e., unresponsiveness to
preload manipulations) remains the dominant pattern in
the obese (regardless of their restraint status), as documented in the various studies cited above, reviewed and
discussed in Ruderman (1986) (but see Footnote 3).
As for normal-weight individuals, there is some evidence
(including both the dieting-overeating correlations in the
present study, not to mention the various experimental lab
studies) that overeating and dieting may co-exist as
behavioral tendencies within the same individual. This is
not to say that all dieters regularly overeat in certain
circumstances. First, the association between dieting and
overeating tendencies is far from perfect; and of course,
some dieters may be sorely tempted to overeat and
nevertheless resist the temptation (Herman & Polivy,
2004). In short, some dieters—especially those in whom
the dieting tendency is stronger than is the overeating
tendency—may well succeed. Such successful dieters are
less likely to be identified by the RS than by the DEBQ and
the TFEQ (Van Strien, 1999).10
It should be noted that it was the disinhibition subscale
of the TFEQ that was solely responsible for the lack of fit
of the model in the sample of female students. Only by
removing this scale from the model was a satisfactory fit
found, indicating that this scale does not match up well
with the present model’s overeating dimension. Although
this finding may seem counterintuitive, it does coincide
with recent results from a confirmatory factor analysis of
the TFEQ (Mazzeo, Aggen, Anderson, Tozzi, & Bulik,
2003), which found that the TFEQ does not display the
three-factor structure that was explicit in the design of the
instrument.
The present study, using large samples, allows us to draw
certain conclusions with reasonable confidence. First, the
RS is clearly related to external measures of dieting,
overeating, and body dissatisfaction. Second, although the
multifactor nature of the RS was confirmed for both the
normal-weight and overweight sample, the RS appears to
capture these constructs differently in overweight than in
normal-weight females. In the normal-weight sample, the
association with the external measures of dieting and body
dissatisfaction was most pronounced, whereas in the
10
Both the DEBQ and TFEQ permit separate assessment of overeating
(disinhibition) and restrained eating. By using a two-factorial classification
including restraint scores and overeating (disinhibition) scores, we should
be able to identify dieters with low susceptibility toward disinhibition
(potentially successful dieters).
ARTICLE IN PRESS
T. van Strien et al. / Appetite 49 (2007) 109–121
overweight sample, the association with the external
measures of overeating was most pronounced, possibly
owing to the RS weight-fluctuation items or to the fact that
overeating leads to overweight. Looked at from another
angle, normal-weight individuals generally accrue high RS
scores because of weight-loss concerns stemming from
body dissatisfaction, whereas overweight individuals may
accrue high RS scores even without being especially
concerned about their weight (or at least without successfully restricting their eating). The result is that the RS
appears not to be a valid measure of restraint in overweight
and obese individuals—they may obtain high restraint
scores without exhibiting restraint in the same way as
normal-weight individuals do—which may in turn explain
the weak association between RS restraint and preloadinduced counterregulation in overweight/obese experimental participants. In contrast, in normal-weight samples, the
RS seems to be valid in that it selects dieters with high a
chance of displaying preload-induced counterregulation—
unsuccessful dieters.
References
Ard, J. D., Desmond, R. A., Allison, D. B., & Conway, J. M. (2006).
Dietary restraint and disinhibition do not affect accuracy of 24-hour
recall in a multiethnic population. Journal of the American Dietetic
Association, 106, 434–437.
Allison, D. B., Kalinsky, L. B., & Gorman, B. S. (1992). A comparison of
the psychometric properties of three measures of dietary restraint.
Psychological Assessment, 4, 391–398.
Arbuckle, J.L., & Wothke, W. (1999). Amos 4.0 user’s guide. Chicago:
Smallwaters Corporation.
Blanchard, F. A., & Frost, R. O. (1983). Two factors of restraint: Concern
for dieting and weight fluctuation. Behaviour Research and Therapy,
21, 259–267.
de Castro, J. M. (1995). The relationship between cognitive restraint to the
spontaneous food and fluid intake of free-living humans. Physiology
and Behavior, 57, 287–295.
Charnock, D. J. (1989). A comment on the role of dietary restraint in the
development of bulimia nervosa. British Journal of Clinical Psychology, 28, 329–339.
Field, A. E., Manson, J. E., Laird, N., Williamson, D. F., Willett, W. C.,
& Colditz, G. A. (2004). Weight cycling and the risk of developing type
2 diabetes among adult women in the United States. Obesity Research,
12, 267–274.
French, S. A., Jeffery, R. W., & Wing, R. R. (1994). Food intake and
physical activity: A comparison of three measures of dieting. Addictive
Behaviors, 19, 401–409.
Ganley, R. M. (1988). Emotional eating and how it relates to dietary
restraint, disinhibition and perceived hunger. International Journal of
Eating Disorders, 7, 635–647.
Garner, D. M. (1991). Eating disorder inventory—2 manual. Odessa, FL:
Psychological Assessment Resources.
Green, M. W., Rogers, P. J., Elliman, N. A., & Gatenby, S. J. (1994).
Impairment of cognitive performance associated with dieting and high
levels of dietary restraint. Physiology and Behavior, 55, 447–452.
Hawkins, R. C., & Clement, P. F. (1980). Development and construct
validation of a self-report measure of binge eating tendencies. Addictive
Behaviors, 5, 219–226.
Heatherton, T. F., Herman, C. P., & Polivy, J. (1991). Effects of physical
threat and ego threat on eating behavior. Journal of Personality and
Social Psychology, 60, 138–143.
119
Heatherton, T. F., Herman, C. P., Polivy, J., King, G. A., & McGree, S.
T. (1988). The (mis)measurement of restraint: An analysis of
conceptual and psychometric issues. Journal of Abnormal Psychology,
97, 19–28.
Heatherton, T. F., Polivy, J., & Herman, C. P. (1991). Restraint, weight
loss, and variability of body weight. Journal of Abnormal Psychology,
100, 78–83.
Herman, C. P., & Mack, D. (1975). Restrained and unrestrained eating.
Journal of Personality, 43, 647–660.
Herman, C. P., & Polivy, J. (1984). A boundary model for the regulation
of eating. In A. J. Stunkard, & E. Stellar (Eds.), Eating and its
Disorders (pp. 141–156). New York: Raven Press.
Herman, C. P., & Polivy, J. (2004). The self-regulation of eating:
Theoretical and practical problems. In R. F. Baumeister, & K. D.
Vohs (Eds.), Handbook of self-regulation: Research, theory and
applications (pp. 492–508). New York: Guilford Press.
Herman, C.P. & Polivy, J. (in press). Norm-violation, norm-adherence,
and overeating. Collegium Anthropologicum.
Herman, C. P., Polivy, J., & Esses, V. M. (1987). The illusion of counterregulation. Appetite, 9, 161–169.
Herman, C. P., Polivy, J., & Leone, T. (2005). The psychology of
overeating. In D. Mela (Ed.), Food, diet and obesity (pp. 115–136).
Cambridge, UK: Woodhead Publishing.
Herman, C. P., Polivy, J., Pliner, P., Threlkeld, J., & Munic, D. (1978).
Distractibility in dieters and nondieters: An alternative view of
externality. Journal of Personality and Social Psychology, 36,
536–548.
Herman, C.P., Polivy, J., & Van Strien, T. (2006). Counter-regulation
present and absent: The pursuit of an elusive phenomenon. Manuscript in preparation.
Hibscher, J. A., & Herman, C. (1977). Obesity, dieting, and the expression
of obese characteristics. Journal of Comparative and Physiological
Psychology, 91, 374–380.
Hill, A. J. (2004). Does dieting make you fat? British Journal of Nutrition,
92(Suppl. 1), S15–S18.
Hu, L. t., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in
covariance structure analysis: Conventional criteria versus new
alternatives. Structural Equation Modeling, 6, 1–55.
Hill, A. J., & Robinson, A. (1991). Dieting concerns have a functional
effect on the behaviour of nine-year-old girls. British Journal of Clinical
Psychology, 30, 265–267.
Hyland, M. E., Irvine, S. H., Thacker, C., Dann, P. L., & Dennis, I.
(1989). Psychometric analysis of the Stunkard-Messick Eating Questionnaire (SMEQ) and comparison with the Dutch Eating Behavior
Questionnaire (DEBQ). Current Psychology: Research and Reviews, 8,
228–233.
Jansen, A., Oosterlaan, J., Merckelbach, H., & Van den Hout, M. A.
(1988). Nonregulation of food intake in restrained, emotional, and
external eaters. Journal of Psychopathology and Behavioral Assessment,
10, 345–354.
Johnson, F., & Wardle, J. (2005). Dietary restraint, body dissatisfaction
and psychological distress: A prospective analysis. Journal of Abnormal
Psychology, 114, 119–125.
Klesges, R. C., Isbell, T. R., & Klesges, L. M. (1992). Relationship
between dietary restraint, energy intake, physical activity, and body
weight: A prospective analysis. Journal of Abnormal Psychology, 101,
668–674.
Klesges, R. C., Klem, M. L., & Bene, C. R. (1989). Effects of dietary
restraint, obesity, and gender on holiday eating behavior and weight
gain. Journal of Abnormal Psychology, 98, 499–503.
de Lauzon-Guillain, B., Basdevant, A., Romon, M., Karlsson., J., Borys,
J., Charles, M. A. & the FLVS Study Group. (2006). Is restrained
eating a risk factor for weight gain in a general population? American
Journal of Clinical Nutrition, 83, 132–138.
Laessle, R. G., Tuschl, R. J., Kotthaus, B. C., & Prike, K. M. (1989a).
A comparison of the validity of three scales for the assessment
of dietary restraint. Journal of Abnormal Psychology, 98,
504–507.
ARTICLE IN PRESS
120
T. van Strien et al. / Appetite 49 (2007) 109–121
Laessle, R. G., Tuchl, R. L., Kotthaus, B. C., & Pirke, K. M. (1989b).
Behavioural and biological correlates of dietary restraint in normal
life. Appetite, 12, 83–94.
Lowe, M. R. (1984). Dietary concern, weight fluctuation and weight
status: Further explorations of the Restraint Scale. Behaviour Research
and Therapy, 22, 243–248.
Lowe, M. R. (1993). The effects of dieting on eating behavior: A threefactor model. Psychological Bulletin, 114, 100–121.
Lowe, M. R. (2002). Dietary restraint and overeating. In C. G. Fairburn,
& K. D. Brownell (Eds.), Eating disorders and obesity (pp. 88–92). New
York: Guilford.
Lowe, M. R., Foster, G. D., Kerzhnerman, I., Swain, R. M., & Wadden,
T. A. (2001). Restrictive dieting vs. undieting: Effects on eating
regulation in obese clinic attenders. Addictive Behaviors, 26,
253–266.
Lowe, M. R., & Kleifield, E. I. (1988). Cognitive restraint, weight
suppression, and the regulation of eating. Appetite, 10, 159–168.
Lowe, M. R., & Levine, A. S. (2005). Eating motives and the controversy
over dieting: Eating less than needed versus less than wanted. Obesity
Research, 13, 797–806.
Lowe, M. R., & Timko, C. A. (2004). What a difference a diet makes:
toward an understanding of differences between restrained dieters and
restrained nondieters. Eating Behaviors, 4, 199–208.
Lowe, M. R., Whitlow, J. W., & Bellwoar, V. (1991). Eating regulation:
The role of restraint, dieting, and weight. International Journal of
Eating Disorders, 10, 461–471.
Masheb, R. M., & Grilo, C. M. (2002). On the relation of flexible and rigid
control of eating to body mass index and overeating in patients with
binge eating disorder. International Journal of Eating Disorders, 31,
82–91.
Mazzeo, S. E., Aggen, S. H., Anderson, C., Tozzi, F., & Bulik, C. M.
(2003). Investigating the structure of the Eating Inventory (ThreeFactor Eating Questionnaire): A confirmatory approach. International
Journal of Eating Disorders, 34, 255–264.
McCann, K. L., Perri, M. G., Nezu, A. M., & Lowe, M. R. (1992). An
investigation of counterregulatory eating in obese clinic attenders.
International Journal of Eating Disorders, 12, 161–169.
Nisbett, R. E. (1972). Hunger, obesity, and the ventromedial hypothalamus. Psychological Review, 79, 433–453.
Ouwens, M. A., Van Strien, T., & Van der Staak, C. P. F. (2003a).
Absence of a disinhibtion effect of alcohol on food consumption.
Eating Behaviors, 4, 323–332.
Ouwens, M. A., Van Strien, T., & Van der Staak, C. P. F. (2003b).
Tendency toward overeating and restraint as predictors of food
consumption. Appetite, 40, 291–298.
Overduin, J., & Jansen, A. (1996). A new scale for use in non-clinical
research into disinhibitive eating. Personality and Individual Differences, 20, 669–677.
Polivy, J., Heatherton, T. F., & Herman, C. P. (1988). Self-esteem,
restraint, and eating behavior. Journal of Abnormal Psychology, 97,
354–356.
Polivy, J., & Herman, C. P. (1985). Dieting and binging: A causal analysis.
American Psychologist, 40, 193–201.
Polivy, J., & Herman, C. P. (1993). Etiology of binge eating: Psychological
mechanisms. In C. G. Fairburn, & G. T. Wilson (Eds.), Binge eating:
Nature, assessment and treatment (pp. 173–205). New York: Guilford
Press.
Presnell, K., & Stice, E. (2003). An experimental test of the effect of
weight-loss dieting on bulimic pathology: tipping the scales in a
different direction. Journal of Abnormal Psychology, 112, 166–170.
Rodin, J. (1981). Psychological factors in obesity. In P. Björntorp, M.
Cairella, & A. N. Howard (Eds.), Recent advances in obesity research,
Vol. III (pp. 106–123). London: John Libbey & Co.
Ruderman, A. J. (1986). Dietary restraint: A theoretical and empirical
review. Psychological Bulletin, 99, 247–262.
Ruderman, A. J., & Christensen, H. (1983). Restraint theory and its
applicability to overweight individuals. Journal of Abnormal Psychology, 92, 210–215.
Ruderman, A. J., & Wilson, G. (1979). Weight, restraint, cognitions, and
counterregulation. Behaviour Research and Therapy, 17, 581–590.
Schachter, S., Goldman, R., & Gordon, A. (1968). Effects of fear, food
deprivation, and obesity on eating. Journal of Personality and Social
Psychology, 10(2), 91–97.
Schoemaker, C., Van Strien, T., & Van der Staak, C. (1994). Validation of
the eating disorders inventory in a nonclinical population using
transformed and untransformed responses. International Journal of
Eating Disorders, 15, 387–393.
Shearin, E. N., Russ, M. J., Hull, J. W., Clarkin, J. F., & Smith, G. P.
(1994). Construct validity of the Three-Factor Eating Questionnaire:
Flexible and rigid control subscales. International Journal of Eating
Disorders, 16, 187–198.
Smith, C. F., Williamson, D. A., Bray, G. A., & Ryan, D. H. (1999).
Flexible vs. rigid dieting strategies: Relationship with adverse
behavioral outcomes. Appetite, 32, 295–305.
Stice, E., Fisher, M., & Lowe, M. R. (2004). Are dietary restraint scales
valid measures of acute dietary restriction? Unobtrusive observational
data suggest not. Psychological Assessment, 16, 51–59.
Stice, E., Ozer, S., & Kees, M. (1997). Relation of dietary restraint to
bulimic symptomatology: The effects of the criterion confounding of
the restraint scale. Behaviour Research and Therapy, 35, 145–152.
Stice, E., Presnell, K., Lowe, M. R., & Burton, E. (2006). Validity of
dietary restraint scales: Reply to van Strien et al. Psychological
Assessment, 18, 95–99.
Stubbs, R. J., Johnstone, A. M., O’Reilly, L. M., & Poppitt, S. D. (1998).
Methodological issues relating to the measurement of food, energy and
nutrient intake in human laboratory-based studies. Proceedings of the
Nutrition Society, 57, 357–372.
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.
Tabachnick, B. G., Fidell, L. S., (2001). Using multivariate statistics.
Boston, MA: Allyn and Bacon.
Timko, C., & Perone, J. (2005). Rigid and flexible control of eating
behaviour in a college population. Eating Behaviours, 6, 119–125.
Tuschl, R. J., Platte, P. P., Laessle, R. G., Stichler, W., & Pirke, K. M.
(1990). Energy expenditure and everyday eating behaviour in healthy
young women. American Journal of Clinical Nutrition, 52, 81–86.
Van Strien, T. (1996). On the relationship between dieting and obese and
bulimic eating patterns. International Journal of Eating Disorders, 19,
83–93.
Van Strien, T. (1999). Success and failure in the measurement of restraint:
Notes and data. International Journal of Eating Disorders, 25, 441–449.
Van Strien, T. (2002a). Dutch Eating Behaviour Questionnaire: Manual.
London: Harcourt Assessment.
Van Strien, T. (2002b). Eating Disorder Inventory II: Handleiding
Nederlandse versie [Eating Disorder Inventory II:Dutch Manual].
Lisse: Swets & Zeitlinger.
Van Strien, T. (2005). Nederlandse Vragenlijst voor Eetgedrag 2005.
Handleiding en Verantwoording [Manual of the Dutch Eating Behavior
Questionnaire 2005]. Amsterdam: Boom test uitgevers.
Van Strien, T., Breteler, M. H. M., & Ouwens, M. A. (2002). Restraint
Scale, its sub-scales concern for dieting and weight fluctuation.
Personality and Individual Differences, 33, 791–802.
Van Strien, T., Cleven, A., & Schippers, G. (2000). Restraint, tendency
toward overeating and ice cream consumption. International Journal of
Eating Disorders, 28, 333–338.
Van Strien, T., Engels, R. C. M. E., Van Leeuwe, J., & Snoek, H. M.
(2005). The Stice model of overeating: tests in clinical and non-clinical
samples. Appetite, 45, 205–213.
Van Strien, T., Engels, R. C. M. E., Van Staveren, W. A., & Herman, C.
P. (2006). The validity of dietary restraint scales: A response to Stice,
Fisher and Lowe (2004). Psychological Assessment, 18, 95–99.
Van Strien, T., Frijters, J. E., Bergers, G. P., & Defares, P. B. (1986a). The
Dutch Eating Behavior Questionnaire (DEBQ) for assessment of
restrained, emotional, and external eating behavior. International
Journal of Eating Disorders, 5, 295–315.
ARTICLE IN PRESS
T. van Strien et al. / Appetite 49 (2007) 109–121
Van Strien, T., Frijters, J. E., Van Staveren, W. A., Defares, P. B., &
Deurenberg, P. (1986b). The predictive validity of the Dutch
Restrained Eating Scale. International Journal of Eating Disorders, 5,
747–755.
Van Strien, T., & Ouwens, M. A. (2003a). Counterregulation in female
obese emotional eaters: Schachter, Goldman, and Gordon’s (1968) test
of psychosomatic theory revisited. Eating Behaviors, 3, 329–340.
Van Strien, T., & Ouwens, M. (2003b). Validation of the Dutch EDI-2 in
one clinical and two nonclinical populations. European Journal of
Psychological Assessment, 19, 66–84.
Van Strien, T., & Ouwens, M. A. (2007). Effects of distress, alexithymia
and impulsivity on eating. Eating Behaviors, in press.
Wardle, J. (1980). Dietary restraint and binge eating. Behavioural Analysis
and Modification, 4, 201–209.
121
Wardle, J. (1987). Eating style: A validation study of the Dutch Eating
Behaviour Questionnaire in normal subjects and women with eating
disorders. Journal of Psychosomatic Research, 31, 161–169.
Wardle, J., & Beales, S. (1987). Restraint and food intake: An
experimental study of eating patterns in the laboratory and in normal
life. Behaviour Research and Therapy, 25, 179–185.
Wardle, J., & Beales, S. (1988). Control and loss of control over eating:
An experimental investigation. Journal of Abnormal Psychology, 97,
35–40.
Wardle, J., Marsland, L., Sheikh, Y., Quinn, M., Fedoroff, I., & Ogden, J.
(1992). Eating style and eating behaviour in adolescents. Appetite, 18,
167–183.
Westenhoefer, J. (1991). Dietary restraint and disinhibition: Is restraint a
homogeneous construct? Appetite, 16, 45–55.