Motivations to eat: Scale development
M. Lynne Cooper,
and Austin Albino
Channing Laboratory, Brigham & WomenÕs Hospital/Harvard Medical School;
Harvard School of Public Health, 181 Longwood Avenue, Boston, MA, USA
Department of Psychology, 105 McAlester Hall, University of Missouri-Columbia,
Columbia, MO 65211, USA
Department of Educational and Counseling Psychology, 16 Hill Hall, University of Missouri-Columbia,
Columbia, MO 65211-2130, USA
Department of Psychological Sciences, 200 S. 7th St., Rm 124, University of Missouri-Columbia,
Columbia, MO 65211, USA
Objective: To validate a measure of psychological motivations to eat based on a four-cat-
egory model of motivations for alcohol use (Cooper, 1994). Motivations speciﬁed by this mod-
el are: to cope with negative aﬀect, to be social, to comply with othersÕexpectations, and to
Method: In Study 1, 40 respondents were queried in an open-ended format about their rea-
sons for eating; responses were content-coded to determine if they ﬁt into the four theorized
categories. In Study 2, an item pool was generated based on responses from Study 1, and ran-
dom halves of a sample of 812 college students were used to test and then validate the hypoth-
esized factor structure.
Results: As expected, the ﬁnal inventory yielded the four theorized categories. The factor
structure was generally invariant across gender, and the resulting Motivations to Eat subscales
uniquely predicted restrictive eating, bingeing, and purging.
Discussion: Prior eating research has focused mainly on coping and compliance motiva-
tions. The present study identiﬁed four distinct motivations to eat that potentially are impor-
tant for understanding healthy and disordered eating.
Ó2003 Elsevier Science (USA). All rights reserved.
Journal of Research in Personality 37 (2003) 297–318
Corresponding author. Fax: +617-525-2578.
E-mail address: firstname.lastname@example.org (B. Jackson).
0092-6566/03/$ - see front matter Ó2003 Elsevier Science (USA). All rights reserved.
Eating is a physiological requirement, insofar as it provides nutrients for proper
growth and functioning of the major systems of the human body (e.g., cardiovas-
cular, immune, and pulmonary). Even beyond physiological requirements for suste-
nance, psychological motivations play an important role in the initiation of both
healthy and disordered eating. For example, eating can be motivated by both po-
sitive (e.g., happiness) and negative (e.g., anger, depression) emotions (Arnow,
Kenardy, & Agras, 1995; Heatherton & Baumeister, 1991; Macht & Simons,
2000; Sherwood, Crowther, Wills, & Ben-Porath, 2000), as well as by a desire to
nurture oneself (Leham & Rodin, 1989). Eating can also be motivated by internal-
ized social cues, norms, and expectations about food consumption (Leary, Tchividj-
ian, & Kraxberger, 1994; Roth, Herman, Polivy, & Pliner, 2001) or by cultural
practices that strengthen feelings of social connection, such as celebrations and
other social functions involving food (Pliner & Chaiken, 1990). Such motivations
to initiate eating have been the focus of some discussion and empirical research,
but these motivations most often have been discussed in isolation from one
Allison (1994) provided an exhaustive review of existing inventories pertaining to
eating behaviors and eating disorders; none of the over 100 inventories reviewed was
designed to assess a range of psychological motivations to eat. Nevertheless, a few
inventories examined one or two aspects of psychological motivations concerning
food-related behaviors. For example, the Emotional Eating Scale (Arnow et al.,
1995) assesses an important psychological motivation for eating: as a response to
negative emotions. Two of the three scales in the Dutch Eating Behavior Question-
naire (van Strien, Frijters, Bergers, & Defares, 1986) tap motivations to eat: Emo-
tional Eating (eating in response to negative emotions) and External Eating
(eating in response to external sensory cues, such as the smell and appearance of
food). Finally, the Eating Inventory (Stunkard & Messick, 1985; cf. Stunkard &
Messick, 1988) has three scales, including one scale (disinhibition of control) that in-
cludes some items reﬂecting psychological reasons to eat. However, the disinhibition
of control scale includes items that reﬂect psychological motivations to eat (e.g.,
‘‘When I feel lonely, I console myself by eating’’) along with other non-motivation
items (e.g., ‘‘My weight has hardly changed at all in the last ten years’’). Thus the
scale does not provide a pure measure of motivations, nor was it intended to. Since
AllisonÕs 1994 book, Steptoe, Pollard, and Wardle published a measure of motiva-
tions that are associated with eating behavior (1995). However, their measure is de-
signed to assess what motivates food choices, once a person already has decided to
eat in the ﬁrst place. In summary, though there are scales related to psychological
motivations to eat, none are theoretically derived for measuring exclusively a range
of motivations to initiate eating.
The relative lack of systematic theoretical and empirical attention to psychologi-
cal motivations in the ﬁeld of eating disorders is surprising, especially when one ex-
amines the literature pertaining to psychological motivations for alcohol use. This
literature is particularly relevant considering the relation between eating disorders
298 B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318
and alcohol use. Speciﬁcally, both alcohol use and eating disorders can be considered
ways of coping with societal, relational, and personal matters in peopleÕs lives (Mintz
& Wright, 1993; cf. Brumberg, 2000). In this view, alcohol and food are both thought
to serve similar functions for an individual (Brisman & Seigel, 1984; cf. Snyder &
Cantor, 1998). The psychological structure of motivations in both domains might
have common features, even though the behaviors themselves are quite distinct
(cf. Cooper, Agocha, & Sheldon, 2000). Whereas the literature about psychological
motivations to eat is quite small (Arnow et al., 1995; Stunkard & Messick, 1985; van
Strien et al., 1986), there is considerable research on psychological motivations for
Cox and Klinger (1988, 1990) proposed a framework for understanding psycho-
logical motivations for alcohol use, and this framework was applied by Cooper and
colleagues (Cooper, 1994; Cooper, Frone, Russell, & Mudar, 1995) to develop a
four-category model of motivation for alcohol use. As stated by Cooper (1994), in
this model, motivations for drinking alcohol
can be meaningfully characterized along two underlying dimensions reﬂecting the valence
(positive or negative) and source (internal or external) of the outcomes an individual hopes
to achieve by drinking. Thus, individuals may drink to obtain a positive outcome...or to
avoid a negative one... Moreover, drinking may be responsive to internal rewards, such
as the manipulation or management of oneÕs own internal emotional state, or to external
rewards, such as social acceptance or approval (p. 118).
Further, by crossing these two dimensions, four categories of motivations for al-
cohol consumption emerge. Speciﬁcally, we see a negatively valenced, internally elic-
ited motivation (to cope with negative aﬀect); a positively valenced, externally
elicited motivation (to be social); a negatively valenced, externally elicited motivation
(to comply with social expectations); and a positively valenced, internally elicited
motivation (to enhance pleasure or positive aﬀect).
The four-category model of drinking motivations has clear parallels to the types
of eating motivations discussed in the literature to date. For example, research has
shown that people regulate their eating behaviors as a way to cope with emotional
distress (Gangley, 1988; Heatherton & Baumeister, 1991; Stice & Agras, 1999; Strau-
man, Vookles, Berenstein, Chaiken, & Higgins, 1991). Whether emotional distress
leads to food restriction or intake depends on an interaction of numerous complex
factors: psychological, historical, cultural, and sociological (Bordo, 1993; Brumberg,
2000; Wolf, 1991). Both eating for social reasons and compliance reasons are inﬂu-
enced by externally derived norms and expectations, though the motivation to eat to
be social is activated only during social occasions. Research suggests a norm of min-
imal eating in othersÕcompany is stronger than the norm of matching the food in-
take levels of oneÕs peers (Roth et al., 2001). Group norms, however, will shape the
baseline that constitutes minimal eating during social gatherings (e.g., in public set-
tings norms for young college women might lead to lesser degrees of food intake
than for their male counterparts; cf. Pliner & Chaiken, 1990; Wolf, 1991). The mo-
tivation to eat to comply—as activated by a set of internalized expectations which
may be chronically primed—can lead to either chronic food intake or restriction
(Fredrickson, Roberts, Noll, Quinn, & Twenge, 1998; Heatherton & Baumeister,
B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318 299
1991; Mintz & Betz, 1998; Noll & Fredrickson, 1998; Strauman et al., 1991; Wolf,
1991). Finally, eating for the sensory pleasure of food has been associated with
binge-eating. Indeed, sensory cues (e.g., delicious aroma or attractive appearance
of foods) have been shown to activate binge-eating episodes (cf. Waters, Hill, &
Waller, 2001), suggesting that pleasure motivations for eating might be the more
proximal cause of over-eating, mediating cue exposure and bingeing. Hence, al-
though the psychological motivations to eat thus far identiﬁed have not been inte-
grated into an overarching theoretical framework, they appear to ﬁt well within
this framework (also see Cooper & Shapiro, 1997, and Cooper, Shapiro, & Powers,
1998, for an elaboration of this framework and its application to other health-
In the current set of studies, we therefore develop and validate an inventory to
assess psychological motivations to eat based on this four-category model (i.e.,
coping, social, compliance, and pleasure motivations). In Study 1, possible scale
items were generated through an open-ended elicitation of 40 respondents, with ad-
ditional face valid items generated by the researchers. In Study 2, the factor struc-
ture of the inventory was determined, and item content was reﬁned, based on data
collected from a separate sample of 812 college students. Given that gender funda-
mentally shapes eating practices and food consumption (Bordo, 1993; Wolf, 1991),
the factor structure was tested for invariance across females and males. As well,
measures to determine several types of validity—convergent, discriminant, and
Based on theoretical and clinical knowledge of food consumption and of the eti-
ology of eating disorders, we formulated general hypotheses about how each of the
motivations would predict three types of eating behaviors: restrictive eating, binge-
ing, and purging. We hypothesized that (a) coping motivations to eat would posi-
tively predict all three types of eating behaviors; (b) social motivations to eat
would negatively predict all three types of eating behaviors; (c) compliance motiva-
tions to eat would positively predict all three types of eating behaviors; and (d) plea-
sure motivations to eat would positively predict binge-eating, negatively predict
restrictive eating, but show no relation to purging.
2. Study 1: Development of the item pool
In Study 1, we used open-ended elicitation procedures for three purposes: (1) to
determine whether the theoretically derived framework ﬁts with peopleÕs own ideas
about reasons, beyond physiological needs, for eating; (2) to assess whether we might
be overlooking an important domain of these motivations; and (3) to obtain preli-
minary data on gender diﬀerences and similarities in these motivations. Eating dis-
orders are far more common among women than men, but are experienced by
both genders (Andersen, 1999; Wolf, 1991; cf. Bordo, 1993). By including both males
and females in the development of this instrument, we can explore whether the four
categories cover the major motivations to eat for both females and males, or if un-
ique domains exist for either gender.
300 B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318
2.1.1. Sample and procedure
A convenience sample of 28 female and 12 male university and community volun-
teers responded in written form to open-ended questions about their motivations to
eat, beyond physical hunger. The mean age was approximately 25 years old. The fol-
lowing question was posed: ‘‘What kinds of feelings, thoughts, or circumstances typ-
ically prompt you (or others) to eat or want to eat (besides simply feeling hungry/
2.2. Results and discussion
Participants generated 203 discrete responses to this question. Narrative responses
were independently coded by two raters into one of the four theoretically expected
motivations to eat: coping, social, compliance, and pleasure. Written descriptions
of each motivation domain were provided in advance to coders to ensure there
was a common understanding of each motivation. Responses judged as not ﬁtting
these dimensions were assigned to an ‘‘other’’ category. Discrepancies between cod-
ers were resolved by discussion until consensus was reached. All psychologically
meaningful reasons for eating responses ﬁt in one of the four categories. Of the re-
sponses generated, 48% were categorized as Coping motivations (e.g., ‘‘stress/de-
pressed’’), 13% were categorized as Social motivations (e.g., ‘‘as a social event’’),
4% were categorized as Compliance motivations (e.g., ‘‘someone saying Ôyou have
to try thisÕ’’), and 16% were categorized as Pleasure motivations (e.g., ‘‘love the taste
of food’’). The 19% of responses that were coded as other motivations were not cen-
trally psychological in nature (e.g., ‘‘hypoglycemic,’’ ‘‘keeping energy level up’’). In
sum, the four-category model seemed to ﬁt well with participantsÕself-generated un-
derstandings about motivations to eat.
Both females and males generated responses consistent with the four-category
model. Of all the reasons for eating that females generated, 51% of the responses were
Coping motivations; 12% were Social motivations; 5% were Compliance motivations;
16% were Pleasure motivations; and 15% were Other motivations. Of all the reasons
for eating that males generated, 48% of the responses were Coping motivations; 15%
were Social motivations; 2% were Compliance motivations; 17% were Pleasure moti-
vations; and 29% were Other motivations. The diﬀerences by gender for percentage
generated in each category were not statistically diﬀerent, v2ð4Þ¼6:00, p>:05.
3. Study 2: Development, reﬁnement, and validation of the motivation to eat subscales
3.1.1. Sample and procedure
An initial pool of items assessing each of the four theorized domains was gener-
ated from two sources. Thirty-three non-redundant responses were taken from Study
B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318 301
1 and edited as needed for clarity. Additional items were drawn from the literature to
supplement the Compliance, Social, and Pleasure categories. The resulting pool in-
cluded 39 items: 17 on the Coping subscale, 7 on the Social subscale, 7 on the Com-
pliance subscale, and 8 on the Pleasure subscale.
Participants for the present study were 812 undergraduates who received credit as
part of their introductory psychology course. The mean age of the sample was 18.8
years old, 64% of the sample was female, and 87% was European American.
Items were administered in random order, as part of a larger questionnaire
packet. Respondents rated the relative frequency of eating for each of the speciﬁed
reasons on a scale from 1 (almost never/never) to 5 (almost always/always).
To establish convergent and discriminant validity of the Motivations to Eat sub-
scales, the Emotional Eating Scale (EES; Arnow et al., 1995) and the Dutch Eating
Behavior Questionnaire (DEBQ; van Strien et al., 1986) were included in the ques-
tionnaire packet. The EES is a 25-item scale that assesses eating in response to neg-
ative emotions, and comprises three subscales assessing eating in response to anger/
frustration, anxiety, and depression, with 11, 9, and 5 items, respectively. Items begin
with the stem ‘‘How strong is your desire to eat when you feel...’’. Responses range
from 1 (no desire to eat)to5(an overwhelming urge to eat). In this study, the reliabil-
ity of each subscale was: Anger/Frustration, a¼:89; Depression, a¼:83; Anxiety,
a¼:83. These three subscales were expected to converge with the Coping subscale
in the Motivations to Eat measure. The DEBQ is a 33-item inventory that has three
subscales: Emotional Eating (eating in response to negative emotions), External Eat-
ing (eating in response to sensory cues), and Restrained Eating (how much one re-
strains or limits eating), with 13, 10, and 10 items, respectively. Responses range
from 1 (never)to5(very often). In this study, the reliability of each of the subscales
was: Emotional Eating, a¼:94; External Eating,
a¼:71; Restrained Eating,
a¼:91. Emotional Eating was expected to converge with the Coping subscale in
the Motivations to Eat measure. Though eating can be activated by sensory cues,
the pleasure-seeking motivation may be the more proximal cause of eating. As such,
External Eating was expected to demonstrate convergent validity with the Pleasure
subscale in the Motivations to Eat measure. Restrained Eating was expected to show
discriminant validity with each of the four dimensions of the Motivations to Eat
To assess the concurrent validity of the Motivations to Eat subscales, criterion
measures of three types of eating behaviors were also included. Respondents an-
swered questions about the presence and the lifetime frequency of eating restraint
(i.e., fasting, appetite control pill use, and strict dieting), bingeing, and purging
(i.e., vomiting, laxative use, and diuretic use). Presence was measured by a yes/no re-
sponse indicating whether the respondent had ever engaged in a given behavior. For
Due to a clerical error, seven items on the external eating subscale were missing from the
302 B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318
purging and restraint, which were each assessed by multiple behaviors, a count of yes
responses was made across the relevant subset of behaviors. Thus scores could range
from 0 (no behaviors endorsed) to 3 (three behaviors endorsed) for both restrained
eating and purging. Respondents who had ever engaged in a given behavior were
then asked to indicate on a 1 (1–2 times in your life)to6(more than 20 times) scale
how often they had ever engaged in this behavior. For restrictive eating and purging,
lifetime frequency was determined by taking the average across the speciﬁc acts that
constituted the behavior.
Finally, to rule out the eﬀects of socially desirable responding, we included the 20-
item version of the Marlowe–Crowne Social Desirability Scale (Strahan & Gerbasi,
1972). This scale comprises items that are answered true or false; half of the items are
reverse–scored. Sample items include ‘‘IÕm always willing to admit it when I made a
mistake’’ and ‘‘I like to gossip at times’’ (reverse–scored). In coding this scale, each
item was assigned the value of 0 or 1, and responses were then summed; 1 denotes
that the participant marked the socially desirable response. In the 1972 article, based
on a sample of 500 university students, the mean and standard deviation for this
scale are 14.5 and 5.4, respectively; no range was reported. In the current sample,
M¼12:21, SD ¼2:14, with answers ranging from 5 to 19.
3.2. Results and discussion
3.2.1. Overview of analyses
The analyses were conducted in ﬁve phases. First, the sample was randomly di-
vided into two subsamples (Group 1, n¼404; Group 2, n¼408).
model was reﬁned in Group 1, and the reﬁned model was cross-validated in Group 2.
Means, standard deviations, and as of the resulting subscales were calculated. Sec-
ond, the ﬁnal model was tested for invariance in the factor structure across gender,
using the pooled sample. Third, analyses were conducted to test for mean gender dif-
ferences in each of the four Motivations to Eat subscales. Fourth, convergent and
discriminant validity with selected established measures were examined. Fifth, multi-
ple regression was used to test the concurrent validity of the Motivations to Eat sub-
aa-vis three distinctive types of eating behavior: restrictive eating, bingeing,
3.2.2. Measurement model reﬁnement and cross-validation
Exploratory factor analyses. An exploratory factor analysis was conducted to de-
termine whether the four theorized factors would be extracted. Using data from
Group 1, we extracted factors through principal-axis factoring followed by oblimin
rotation, and examined the pattern matrix. Principal-axis factoring was used because
it has been shown to reproduce the initial correlation matrix more reliably than other
extraction techniques (Snook & Gorsuch, 1989), whereas oblimin rotation (which
Group 1 and Group 2 did not diﬀer in gender composition or on mean levels of age, weight, socially
desirable responding, or any of the individual motivation to eat items.
B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318 303
allows correlated factors) was used because earlier research suggests that eating
motivations are correlated (e.g., van Strien et al., 1986). The resulting solution
was evaluated to determine whether items loaded on their intended factors at or
above .40, and demonstrated purity (Floyd & Widaman, 1995). In this case, items
whose loading on a secondary factor were less than half the value of the loading
on its primary or intended factor were considered pure (i.e., non-complex). For ex-
ample, if an item loaded. 60 on its primary factor and .30 or above on another factor,
it would be considered factorially complex.
Exploratory factor analysis of the 39 items yielded six factors with eigenvalues
greater than one. However, the sixth factor was barely over 1 (1.06), and the scree plot
suggested either 4 or 5 factors. Thus, the analysis was re-run constraining the solution
to ﬁve and four factors, respectively. In the ﬁve-factor solution, 18 eat to cope with
negative emotion items loaded on the ﬁrst factor. All items loaded at or above .40,
and only two of the items failed to meet criteria for non-complexity. Six of the seven
social items loaded on factor two with loadings P:40. However, one of these six so-
cial items failed to meet our criterion for non-complexity, thus yielding ﬁve social
items that met both criteria. The third factor contained the compliance items, all
six of which met our dual criteria (i.e., loaded above .40 and showed purity). The
fourth factor was deﬁned by four of the eight pleasure motivation items, with loadings
P:40. Three pleasure motivation items loaded on factor ﬁve, with loadings P:44.
One of the three pleasure items loading on factor ﬁve cross-loaded on factor four. Fi-
nally, one pleasure item loaded below .40 on both the fourth and ﬁfth factors. In gen-
eral, the items loading on factors four and ﬁve were distinguished by the degree to
which the wording emphasized use of food as a reward or treat vs. eating in order
to experience pleasurable sensations. Constraining the solution to four factors yielded
a structure that was consistent with this conceptualization. Speciﬁcally, the ﬁrst three
factors in the constrained four-factor solution were identical to those obtained in the
ﬁve-factor solution (reﬂecting Coping, Social, and Compliance factors, respectively),
whereas the seven items that had loaded on factors 4 and 5 combined to yield one
Pleasure motivation factor. Examination of individual item loadings on the four fac-
tors showed that all items loaded P:52 on their respective factors, and that all but
one item met our criteria for purity.
Based on the results of these analyses, a subset of ﬁve coping items that met our
dual criteria (i.e., loaded P:40 on its intended factor and had a 2:1 loading ratio on
the primary vs. secondary factors) and reﬂected a range of emotional states were se-
lected for inclusion in the ﬁnal item pool. The ﬁve social items that met our dual cri-
teria were also retained, as were ﬁve of the six compliance items that met these
criteria. Finally, six of the seven pleasure items that met our dual criteria in the con-
strained four-factor solution were retained: three items reﬂecting motivations to eat
related to the pleasurable aspects of food, and three items reﬂecting the use of food
to reward or treat oneself. This ensured the broadest possible coverage of the plea-
sure motivations construct. Thus, a total of 21 items were retained on the four the-
oretically predicted factors.
Conﬁrmatory factor analyses in Group 1. A correlated four-factor model was spec-
iﬁed and tested in Group 1 using conﬁrmatory factor analyses (CFA). The ﬁt of the
304 B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318
model to the data was compared to three alternative models (see Table 1, models
1–4). A one-factor model tested the notion that a single underlying motivation is ad-
equate to account for the pattern of covariation among all of the items. Two alter-
native, correlated two-factor models were also tested: (a) we speciﬁed internal vs.
external factors in which Coping and Pleasure items were constrained to load on
the internal factor and Compliance and Social items were constrained to load on
the external factor; (b) we speciﬁed positive vs. negative reinforcement factors in
which Social and Pleasure items were constrained to load on the positive reinforce-
ment factor and Compliance and Coping items were constrained to load on the neg-
ative reinforcement factor.
All CFAs were performed using the EQS structural equation modeling program
(Bentler, 1995). The variance–covariance matrix served as input. Scaling metrics for
the latent variables were ﬁxed by setting factor variances to 1.0. Following the rec-
ommendation of several authors (Bollen & Long, 1993; Loehlin, 1998), multiple ﬁt
indices are reported for each of the analyses, including the v2/df ratio (Bollen,
1989), the normed ﬁt index (NFI; Bentler & Bonnett, 1980), the comparative ﬁx in-
dex (CFI; Bentler, 1995), and the standardized root-mean-square residual (RMR).
Given that each ﬁt index has diﬀerent limitations, consistency across indices is gen-
erally regarded as the most reliable indicator of goodness of ﬁt (Mulaik et al., 1989).
Although no consensus exists on the exact value of the v2/df ratio needed to indicate
good ﬁt, all recommendations fall in the range of 2:1 to no more than 5:1 (Bollen,
1989; Marsh & Hocevar, 1985). The values for both the NFI and the CFI range from
Development and validation of factor structure across randomly split halves
Model v2df v2/df NFI CFI RMR
1. Original four-factor model: Group 1 749.51183 4.10 .88 .85 .05
2. Final four-factor model: Group 1 512.74164 3.13 .89 .92 .05
3. One-factor model: Group 1 1958.34170 11.52 .56 .59 .09
4. Two-factor modela
(internal vs. external): Group 1
1547.47169 9.16 .66 .68 .10
5. Two-factor modelb
(positive vs. negative): Group 1
1380.96169 8.17 .69 .72 .09
6. Final four-factor model: Group 2 354.77164 2.16 .92 .95 .04
7. Factor pattern invariant model:
1195.56292 4.09 .91 .93 .05c
8. Factor-loading invariant model:
898.35344 2.61 .90 .94 .05c
Note. NFI, Normed Fit Index; CFI, Comparative Fit Index; RMR, standardized root-mean-square
residual. In Group 1, n¼404, and in Group 2, n¼408 (total N¼812).
Compliance and Social items were constrained to load on an external factor; Coping and Pleasure
items were constrained to load on an internal factor.
Social and Pleasure items were constrained to load on a positive factor; Compliance and Pleasure
items were constrained to load on a negative factor.
Fit of the model to the data in Group 1.
Fit of the model to the data in Group 2.
B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318 305
0 to 1, with values from .90 to .95 as acceptable, and higher values indicating good ﬁt
(Bentler & Bonnett, 1980; Bollen & Long, 1993; Hu & Bentler, 1999). The RMR is
the standardized average absolute diﬀerence between the original and the reproduced
matrices (Marsh & Hocevar, 1985). Small values (e.g., .05) indicate minimal discrep-
ancy between the original and the reproduced matrices. RMR values reported here
exclude diagonal elements and as such are more conservative.
Goodness-of-ﬁt information for the four alternative models (tested in Group 1) is
summarized in Table 1, models 1–4. The original 21-item four-factor model did not
ﬁt the data well, as demonstrated by NFI and CFI values of .88 and .85, respectively.
Examination of the modiﬁcation indices (using the LaGrange Multiplier Test; Ben-
tler, 1995) revealed factorial complexity associated with the Pleasure item ‘‘because
eating is pleasurable and enjoyable.’’ For example, it appeared that there would be
substantial improvement of ﬁt by allowing the error variance of this item to correlate
with other error variances. To construct a cleaner scale, and because the Pleasure
scale already had six items and the other scales had ﬁve, we re-speciﬁed the four-fac-
tor model by dropping this one item and proceeding with the ﬁve remaining items on
the Pleasure scale. This modiﬁed four-factor model yielded a more acceptable ﬁt: the
v2/df ratio was 3.13, values for the NFI and CFI were .89 and .92, respectively, and
the RMR was .05.
As shown by the v2-diﬀerence test, the ﬁnal correlated four-factor model ﬁt the
data signiﬁcantly better than did the one-factor model (Model 3–Model 2,
Dv2½6¼1445:60, p<:001), or either of the two-factor models (Model 4–Model 2,
Dv2½5¼1034:73, p<:001 and Model 5–Model 2, Dv2½5¼868:22, p<:001).
3.2.3. Validation in Group 2
To determine whether this modiﬁed factor structure was reliable, we speciﬁed and
tested the invariance of the correlated four-factor structure across the two random
samples. As shown in Table 1, line 6, the correlated four-factor model provided an
acceptably good ﬁt to the data in the second random half of the sample. The v2/df ra-
tio was 2.16, values for the NFI and CFI were .92 and .95, respectively, and the RMR
was .04. To determine whether the speciﬁed model ﬁt equally well across the two
groups, two simultaneous, between-group models were speciﬁed. First, a model
was tested in which a common factor-pattern was speciﬁed across Groups 1 and 2,
but the magnitude of the factor-loadings was allowed to vary. As shown on line 7
of Table 1, the v2/df ratio of 4.09, values for the NFI and CFI of .91 and .93, respec-
tively, and acceptable RMR values indicate that the speciﬁed four-factor model pro-
vided an equally adequate ﬁt to the data across Groups 1 and 2, thus indicating
conﬁgural invariance across subsamples (Widaman & Reise, 1997).
Next, we tested a factor-loading equivalent model in which the factor-pattern as
well as the factor-loadings were constrained to equivalence across Groups 1 and 2
(line 8, Table 1). Examination of the ﬁt indices for this model indicate that the fac-
tor-loading equivalent model also provided an acceptably good ﬁt to the data across
subsamples. The v2/df ratio was 2.61, values for the NFI and CFI were .90 and .94,
respectively, with acceptable RMR values. However, the v2-diﬀerence test between
the factor-pattern only and the factor-pattern with factor-loading equivalent models
306 B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318
was signiﬁcant, Dv2ð52Þ¼297:21, p<:01, indicating that the assumption of com-
plete factor-loading invariance was not supported.
When modiﬁcation indices (using the LaGrange Multiplier Test) for individual
items were examined, results revealed factor-loading invariance for all individual
items across Groups 1 and 2 except for two items (one Coping item: ‘‘because you
feel worthless or inadequate’’ and one Compliance item: ‘‘because you donÕt want
to stand out or be diﬀerent from others who are eating’’). The unstandardized load-
ing for the Coping item in Group 1 was :52 ð:041Þ,p<:001; in Group 2, it was
:64 ð:041Þ,p<:001. The unstandardized loading for the Compliance item in
Group 1 was 1:00 ð:067Þ,p<:001; in Group 2, it was :72 ð:059Þ,p<:001. De-
spite the lack of complete factor-loading invariance, all items loaded signiﬁcantly
on the Coping motivations factor in both groups (tvalues ¼12.61–17.64 and
15.35–20.37 for Group 1 and Group 2, respectively, pvalues <:001), as did all Com-
pliance motivation items in both groups (tvalues ¼11.55–14.89 and 11.89–14.36 for
Group 1 and Group 2, respectively, pvalues <:001). In fact, all items loaded signif-
icantly in both groups for the other two scales as well (Social: tvalues ¼15.45–20.15
and 12.62–18.71 for Group 1 and Group 2, respectively, pvalues <:001; Pleasure: t
values ¼9.63–10.13 and 9.95–10.85 for Group 1 and Group 2, respectively, pvalues
<:001). Internal consistency was also comparable across groups for all four sub-
scales (Coping a¼:88 and .89; Compliance a¼:85 and .84; Social a¼:89 and
.88; Pleasure a¼:82 for both). Given strong evidence for conﬁgural invariance
and evidence of nearly complete factor-loading invariance across the subsamples
for the four-factor model, data were pooled for the remaining analyses.
Table 2 presents the standardized and unstandardized (with standard error) fac-
tor-loadings, for the correlated four-factor model estimated in the entire sample
(N¼812). Descriptive statistics and factor intercorrelations for the resulting scales
are presented in Table 3. As shown in Table 3, the four subscales possess adequate
reliability and show adequate independence among themselves. Interestingly, the
rank order of mean levels of endorsement for each motivation subscale is identical
to what has been observed in the alcohol motivations literature (Cooper, 1994). Spe-
ciﬁcally, Social (M¼2:72, SD ¼:93) and (Pleasure M¼2:31, SD ¼:81) motivations
are endorsed as the most common, followed by Coping (M¼1:65, SD ¼:78) and
Compliance (M¼1:58, SD ¼:70).
3.2.4. Invariance across gender
To determine whether the ﬁnal correlated four-factor model was invariant across
females and males, two simultaneous, between-group models were speciﬁed. First, a
model was tested in which a common factor-pattern was speciﬁed across females and
males, but the magnitude of the factor-loadings was allowed to vary. As shown on
line 3 of Table 4, the v2/df ratio of 2.78, values for the NFI and CFI of .90 and
.93, respectively, and acceptable RMR values indicate that the speciﬁed four-factor
model provided an equally good ﬁt to the data across females and males, thus indi-
cating conﬁgural invariance across subsamples.
Next, a factor-loading equivalent model was tested in which both the factor-pat-
tern and the factor-loadings were constrained to equivalence across females and
B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318 307
Standardized and unstandardized factor loadings and standard errors for the hypothesized four-factor model using the pooled sample (N¼812)
Item Coping Social Compliance Pleasure
SFL USFL SE SFL USFL SE SFL USFL SE SFL USFL SE
9. Because youÕre depressed or sad .75 1.00 NAa
16. Because you feel worthless or inadequate .64 .58 .030
21. As a way to help you cope .88 .99 .037
22. As a way to comfort yourself .82 .94 .039
27. As a way to avoid thinking about something
unpleasant or to distract yourself
.74 .88 .041
7. Because itÕs a special or traditional part
of some social occasion or celebration
.70 .82 .041
29. As a way to enjoy a social gathering .82 .91 .033
31. As a way to celebrate a special occasion
with friends, family, or a loved one
.84 1.00 NAa
32. To be sociable .78 .82 .032
37. To join in a festive occasion .84 .95 .034
34. To keep people from asking questions
about why youÕre not eating
.74 1.00 NAa
35. Because someone pressures you to eat .75 .92 .046
36. Because you feel like you canÕt say ÔnoÕ.75 .95 .048
38. Because you don’twant to stand out or be
diﬀerent from others who are eating
.79 .86 .045
39. To please your mother or someone else
who wants you to eat
.61 .91 .055
3. Because you want to treat yourself .76 1.00 NAa
5. As a reward for having done something
that youÕre proud of or feel good about
.72 .87 .058
6. Because you like to eat .53 .96 .045
10. Because you deserve it .79 .91 .045
19. Because you feel good or are in a good mood .69 .83 .044
Note. Items were preceded by the stem ‘‘How often do you eat.’’ Item numbers reﬂect the order in which items were presented to respondents. SFL,
standardized factor loading, USFL, unstandardized factor loading. All factor loadings are signiﬁcant at p<:001.
NA, Not applicable. For identiﬁcation purposes, the USFL was set to 1.
308 B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318
males. Examination of the ﬁt indices for this model indicated that the factor-loading
equivalent model provided an acceptable ﬁt to the data across females and males
(line 4, Table 4). Values for the NFI and CFI were .89 to .93, respectively, and
RMRs were acceptable (.05 for both groups). The v2-diﬀerence test was not signiﬁ-
cant, Dv2ð16Þ¼1:77, p>:05, indicating that the assumption of complete factor-
loading invariance was supported. Nevertheless, examination of the modiﬁcation
indices (using the LaGrange multiplier test) showed that relaxing the equality con-
straint on one Coping item (‘‘as a way to avoid thinking about something unpleasant
or to distract yourself’’) would signiﬁcantly improve model ﬁt. The unstandardized
loading for this item among females was :87 ð:052Þ,p<:001; among males, it was
1:00 ð:080Þ,p<:001. Not surprisingly, given earlier ﬁndings (see Table 2), all re-
maining items loaded signiﬁcantly on their intended factors among both males
and females. As well, internal consistency was comparable across female and male
respondents on the Coping motivations subscale (a¼:88 and .87, respectively),
and on the Social (a¼:90 and .86, respectively), Compliance (a¼:85 and .83, re-
spectively), and Pleasure (a¼:82 and .83, respectively) motivations subscales.
These analyses indicate a largely, but not entirely, homogeneous solution. Social,
Compliance, and Pleasure motivations met the strict statistical assumptions of
Scale statistics and factor correlations
Scale MSDaFactor correlations
1. Coping 1.65 .78 .88 – .40 .48 .54
2. Social 2.72 .93 .88 – .48 .65
3. Compliance 1.58 .70 .84 – .41
4. Pleasure 2.31 .81 .82 –
Note. Data are from total sample (N¼812). Possible scale means range from 1 (almost never/never) to
5 (almost always/always), in response to item stems that began ‘‘How often do you eat....’’ Means,
standard deviations, and coeﬃcient as are based on scale scores derived from manifest variables; inter-
correlations are among latent factors.
Examination of factor invariance across gender
Model v2df v2/df NFI CFI RMR
1. Female (n¼520) 534.66164 3.26 .91 .93 .05
2. Male (n¼278) 360.04164 2.20 .88 .93 .05
3. Factor pattern invariant model 911.55328 2.78 .90 .93 .05a.05b
4. Factor-loading invariant model 913.27344 2.65 .89 .93 .05a.05b
5. v2diﬀerence 1.77 16 — — — —
Note. NFI, Normed Fit Index; CFI, Comparative Fit Index; RMR, standardized root-mean- square
residual. Dashes indicate data not applicable.
Fit of the model to the data among Females.
Fit of the model to the data among Males.
B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318 309
complete factor-loading invariance for all items across gender groups, whereas all
but one Coping motivations item met the strict statistical assumptions of complete
factor-loading invariance. In summary, these data indicate factor-pattern invariance
for all scales across females and males and complete factor-loading invariance for all
but a single item. Given arguments that factor-pattern invariance is crucial, and that
more stringent forms of invariance cannot be reasonably expected between non-ran-
domly formed groups (Horn, McArdle, & Mason, 1983), our results suggest that
more than adequate similarities exist to permit meaningful cross-group comparisons.
3.2.5. Mean diﬀerences across gender in motivations to eat
To determine whether the Motivations to Eat subscales varied as a function of
gender, we conducted a one-way multivariate analysis of variance (MANOVA).
As shown in Table 5, results revealed that there were no gender diﬀerences in mean
levels of Social motivations (F¼:81, p>:10), Compliance motivations (F¼1:04,
p>:10), and Pleasure motivations (F¼1:55, p>:10). However, there were signif-
icant gender diﬀerences in Coping motivations (F¼24:60, p<:001). Female re-
spondents were signiﬁcantly more likely than male respondents to endorse eating
to cope with emotional distress. Finally, supplementary analyses showed that this
diﬀerence was not due to age, weight, or socially desirable responding. Indeed, when
age, weight, and socially desirable responding were entered as covariates, the eta-
squared for Coping motivations increased, from about 5% of the variance to just
3.2.6. Convergent and discriminant validity
The data in Table 6 illustrate that the Motivations to Eat subscales correlated in
expected ways with the established measures included to assess convergent and dis-
criminant validity. The Coping motivations subscale showed the strongest signiﬁcant
and positive correlations with each of the EES subscales (Anger/Frustration, Anxi-
ety, and Depression) and with the DEBQ Emotional Eating subscale, as expected.
Also as expected, Pleasure motivations were more strongly related to the DEBQ Ex-
ternal Eating Subscale than to any of the other scales. The modest correlations be-
tween Compliance motivations and the included criterion scales, as well as between
Social motivations and the criterion scales, indicate that neither Compliance nor So-
cial motivations are well-represented by existing measures. Finally, and as expected,
Unweighted means for four motivations to eat subscales, by gender
Gender Coping Social Compliance Pleasure
M SD M SD SD SD M SD
Female (n¼520) 1.78 .82 2.74 .96 1.60 .73 2.33 .79
Male (n¼278) 1.40 .61 2.66 .89 1.54 .66 2.28 .84
g2.053.002 .001 .001
Note. WilksÕlambda ¼.94; multivariate F:ð4;807Þ¼13:81, p<:001.
310 B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318
the DEBQ Restrained Eating subscale correlated only modestly (rs<:30) with the
Motivations to Eat subscales, showing discriminant validity. Indeed, we would not
expect motivations to eat to be the same as motivations not to eat.
Finally, socially desirable responding did not correlate substantially with any of
the Motivations to Eat subscales, suggesting that responses to the subscales are more
than simply a function of what participants think they ought to say about their rea-
sons for eating. Overall, these data suggest that the eating motivations subscales
show both convergent and discriminant validity.
3.2.7. Motivations to eat as predictors of eating-related behavior
To establish concurrent validity of the Motivations to Eat subscales and the utility
of distinguishing among these motivations, we tested whether the subscales were dif-
ferentially related to three types of eating behaviors (restrictive eating, bingeing, and
purging). A series of six hierarchical multiple regression analyses was conducted. On
the ﬁrst step, gender was entered and on the second step the set of four motivations to
eat was entered.
For each of the three types of eating behaviors the following two
indicators were examined as dependent variables: ever engaged in the behavior,
and lifetime frequency among those who ever engaged in the behavior. These indica-
tors were chosen so that we could examine the utility of the eating motivations sub-
scales in predicting both initial engagement and degree of involvement in the eating
For dependent measures that were continuous, Ordinary Least Squares (OLS) estimation was used.
For the dichotomous dependent measure (ever binged), analyses were conducted with both logistic and
OLS regression procedures. Although the use of dichotomous dependent variables violates assumptions
underlying OLS estimation (see Neter, Kutner, Nachtsheim, & Wasserman, 1996), comparison of results
from the two procedures showed no substantive diﬀerences. Thus, to simplify reporting and maintain
comparability across dependent measures, results are tabled for the OLS regression procedures only.
Correlations among motivations to eat subscales and convergent/divergent criterion scales
Coping Social Compliance Pleasure
Convergent/Divergent Criterion Scales
EES, Anger/Frustration Subscale 72 31 32 34
EES, Anxiety Subscale 67 29 34 34
EES, Depression Subscale 65 27 30 33
DEBQ, Emotional Eating Subscale 68 22 26 31
DEBQ, External Eating Subscale 34 33 17 47
DEBQ, Restrained Eating Subscale 30 08 21 03
M-C Social Desirability 08 02 05 12
Gender (1 ¼male, 2 ¼female) 23 04 04 00
Note. N¼812, except n¼810 for Social Desirability. Except for Gender, all measures are scored so
that higher numbers indicate more of the measured construct. Decimals are omitted. All correlations
P:07 are signiﬁcant at p<:05. EES, Emotional Eating Scale. DEBQ, Dutch Eating Behavior Ques-
tionnaire. MC, Marlowe–Crowne.
B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318 311
As shown in Table 7, the four Motivations to Eat subscales accounted for approx-
imately 3 to 13% of the variance across restrictive eating, bingeing, and purging. Ex-
amination of the individual beta weights indicated that each of the four motivations
was associated with a unique pattern of eating behavior.
Restrictive eating. As shown in Table 7, the four Motivations to Eat subscales ac-
counted for about ﬁve to six percent of the variance in restrictive eating, after ac-
counting for gender. In general, Coping and Compliance motivations each
signiﬁcantly and positively predicted more restrictive eating for both indices. The ex-
ception was that Coping did not predict lifetime frequency of restrictive eating. Also
as expected, both Pleasure and Social motivations were signiﬁcantly and negatively
related to more restrictive eating, though Pleasure did not predict lifetime frequency.
Binge eating. As shown in Table 7, the four Motivations to Eat subscales ac-
counted for about 13% of variance in the dichotomous binge eating measure, after
accounting for gender. Coping, Compliance, and Pleasure motivations were each sig-
niﬁcantly and positively related to more binge eating. Social motivations were signif-
icantly and negatively related to more binge eating. Lifetime frequency of bingeing
was not related to any of the Motivations to Eat subscales. This suggests that the
motivations are primarily useful in distinguishing between those who ever vs. never
engaged in the behavior, and that they, at least in this sample, do not help us further
understand the degree of involvement in this behavior.
Purging. As shown in Table 7, the Motivations to Eat subscales accounted for
about three percent of the variance in the occurrence of purging behaviors, after ac-
Multiple regression analyses predicting eating behaviors from four motivations to eat subscales
Variable DR2Coping Social Compliance Pleasure
Count ever (n¼811) .064 .15 ).09.24 ).05
.051 .16 ).25þ.42 .04
Ever binge (n¼810) .127 .14 ).05 .08 .08
.019 ).02 ).12 .19 .27
Count ever (n¼811) .031 .06).06.10 .01
.078 .01 .18 .40þ.13
Note. All analyses are after controlling for gender. Lifetime frequency variables include only those
respondents who have ever engaged in the behavior.
Each indicator of ‘‘Restrictive eating’’ is a composite measure of fasting, appetite control pill use, and
going on a strict diet.
Each indicator of ‘‘Purging’’ is a composite measure of vomiting, laxative use, and diuretic use.
312 B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318
counting for gender. Coping and Compliance motivations were signiﬁcantly and pos-
itively related to purging, as expected. In contrast, but also as expected, Social moti-
vations were signiﬁcantly and negatively related to purging, whereas Pleasure
motivations were unrelated. However, generally the motivations did not predict life-
time frequency of purging; Compliance did so only marginally so. As with bingeing,
this suggests that the motivations are primarily useful in distinguishing between those
who ever vs. never engaged in the behavior, and that they, at least in this sample, do
not really help us further understand the degree of involvement in this behavior.
3.2.8. Incremental validity
A series of regression equations was estimated to determine whether the four Mo-
tivations to Eat subscales would predict these criterion-related behaviors above and
beyond existing measures. Speciﬁcally, on the ﬁrst step, the EES subscales, the
DEBQ subscales, and gender were entered, and on the second step, the set of four
Motivations to Eat subscales was entered. Because the Emotional Eating subscale
of the DEBQ and the EES subscales assess the same general content (the DEBQ
EE subscale correlates with EES anger/frustration, r¼:74; EES anxiety, r¼:67;
and EES depression, r¼:64, ps<:001), to reduce multicollinearity, only the EES
scales were included in the reported analyses.
As shown in Table 8, each motivation
to eat predicted the given eating behavior, in an overall pattern similar to in Table 7,
above and beyond the EES subscales and the DEBQ scale. There was one important
exception: including these measures on the ﬁrst step eliminated or substantially re-
duced the predictive validity of Coping motivations. This was expected, though, gi-
ven the substantial overlap of the Coping motivations measure with the EES
subscales and DEBQ aﬀect-related subscale noted in the convergent validity section.
Notably even with the DEBQ External Eating scale in the equation, Pleasure moti-
vations still predicted ever-bingeing. Thus, our measure appears to be an eﬀective
shorter substitute for these longer measures.
In sum, each eating motivation independently is associated with a unique pattern
of eating behaviors, in expected ways, and predicts these behaviors above and be-
yond other established measures of eating motivations.
3.2.9. Eﬀects of motivations to eat by gender
To determine if the eﬀects of motivations to eat diﬀer signiﬁcantly between fe-
males and males, all six hierarchical multiple regression analyses from Table 7 were
estimated with the addition of Gender X Motivation interaction terms on the last
step. To reduce multicollinearity among the interaction terms and their constituent
variables, all variables were centered before computing and entering their interac-
tions (Aiken & West, 1991). To probe signiﬁcant interactions, simple slopes were cal-
culated from the overall regression equation to illustrate the relationship between a
We also tested an alternative model in which the DEBQ Emotional Eating subscale was included
instead of the EES subscales. Results paralleled the ﬁndings reported in Table 8.
B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318 313
given eating motivation and criterion variable among females and males separately
(Aiken & West, 1991).
Six blocks of Gender X Motivation interaction terms (corresponding to each of
the six dependent measures in Table 7) were tested. Of these, the only signiﬁcant in-
teraction was Coping diﬀerentially predicting lifetime frequency of binge eating by
gender. Speciﬁcally, for males but not for females, the stronger the Coping Motiva-
tion, the lower the lifetime frequency of binge eating in the full sample (for males,
b¼:29, p<:05; for females, b¼:11, p>:05). In sum, the analysis of interaction
eﬀects suggests that eating motivation eﬀects are largely invariant between females
and males, with a possible exception in predicting some aspects of binge eating.
4. General discussion
We developed and tested a four-category model of motivations to eat, extending
parallel research done in the domain of alcohol use motivations (Cooper, 1994). As
expected, analyses indicated the presence of four correlated factors that comprise
motivations to eat: Coping, Social, Compliance, and Pleasure. This model provided
an equally good ﬁt for men and for women, as shown by tests of factorial invariance
and by demonstrating good reliability across gender groups. Analysis of mean diﬀer-
ences across groups suggests that, while there are no gender diﬀerences in self-
Multiple regression analyses predicting eating behaviors from four eating motivations, alternative analyses
Variable DR2Coping Social Compliance Pleasure
Count ever (n¼811) .037 .09 ).08.22 ).05
.034 .11 ).23þ.38 ).02
Ever binge (n¼810) .028 .07).05 .08 .05
.010 ).15 ).10 .20 .17
Count ever (n¼811) .016 .02 ).05.09 .01
.073 ).15 .27 .30 .21
Note. All analyses are after controlling for EES Anger/Frustration, EES Anxiety, EES Depression,
DEBQ External Eating, & Gender. Lifetime frequency variables include only those respondents who have
ever engaged in the behavior.
Each indicator of ‘‘Restrictive eating’’ is a composite measure of fasting, appetite control pill use, and
going on a strict diet.
Each indicator of ‘‘Purging’’ is a composite measure of vomiting, laxative use, and diuretic use.
314 B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318
reported mean levels of Social, Compliance, and Pleasure motivations to eat (even
when controlling for the potential confounds of age, weight, and socially desirable
responding), women endorse higher levels of Coping motivations than do men.
The validity of the correlated four-category model was further reinforced by the
fact that each eating motivation was associated with a unique pattern of eating be-
haviors, despite overlap among Motivations to Eat subscales. As hypothesized, Cop-
ing and Compliance motivations to eat both positively predicted restrictive eating,
bingeing, and purging. Given that these are both aversive motivations, we would ex-
pect some similarity between them. However, Coping and Compliance motivations
are diﬀerentiated from each other in that Coping motivations predicted bingeing
more strongly, and Compliance predicted restrictive eating and purging more
strongly (see Table 7). As expected, Pleasure motivations positively predicted binge
eating and negatively predicted restrictive eating, and showed no relation to purging.
Finally, Social motivations negatively predicted restrictive eating and purging, but
positively predicted bingeing.
It seems then that healthy eating is in part the result of meeting the needs of these
four psychological motivations in ways that are not dependent on food for their ful-
ﬁllment, and that disturbed eating results from chronically acting on these motiva-
tions. For our present purposes, the absence of restrictive eating, bingeing, or
purging represents healthy eating. Thus, when a motivation negatively predicts a be-
havior, this points us toward an understanding of the psychological bases of healthy
eating. According to these data, social motivations negatively predict restrictive eat-
ing, bingeing, and purging. We speculate that healthy eating at least in part results
from thoughtful, creative, and ﬂexible responses to social motivations, and disturbed
eating results from a unaware, chronic, and rigid responses to any of the psycholog-
ical motivations to eat (cf. Heatherton & Baumeister, 1991). Moreover, we suggest
that healthy eating requires that the needs corresponding to social motivations are
fulﬁlled in ways that are independent of food alone (Mintz & Wright, 1993; Roth,
1993). We acknowledge that we did not assess important aspects of healthy eating
behaviors, such as avoiding high fat or high sugar foods, or eating very healthy
foods, such as fruits and vegetables (Conner, Norman, & Bell, 2002). Assessing
how the four motivations predict these types of healthy eating behaviors could be
more fully explored in future research.
The Motivations to Eat measure is the ﬁrst validated instrument that is both the-
oretically-based and that examines a range of psychological motivations pertaining
to eating. As expected, the Coping motivation subscale overlaps with other similar
established scales, like the DEBQ and EES subscales. Hence, if an instrument is
needed to distinguish among speciﬁc negative emotions, then researchers should con-
sider using measures such as the EES. If an instrument is needed that provides a
comprehensive measure of motivations that includes a more general Coping motiva-
tions subscale—akin to, but shorter than established scales like the DEBQ Emo-
tional Eating subscales—then our measure should be administered.
Such an instrument has both clinical and research utility. The Motivations to Eat
measure could be used in both individual treatment and prevention eﬀorts of eating
disorders. The measure could be given to clients in treatment for eating disorders, as
B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318 315
well as those identiﬁed as at-risk for eating disorders, to assess the reasons underly-
ing their food intake. Counseling could then revolve around the identiﬁcation of
other ways to meet these needs. The Motivations to Eat measure could also be used
in large-scale research aimed at identifying both risk and protective factors associ-
ated with the development of eating disorders. Studies could be conducted to exam-
ine the role that speciﬁc psychological motivations have in the etiology of eating
disorders, including bulimia, anorexia, and EDNOS (eating disorders not otherwise
speciﬁed). Likewise, it could be used in outcome research on the treatment of eating
disorders (i.e., do motivation scale scores change as a result of treatment?).
Several limitations of the present study should be acknowledged. First, our samples
were limited to college students. As well, while we speculate that our eating motivations
measure will have clinical relevance, we did not use indices of eating disorders per se,
nor did we examine the subscales in samples of individuals diagnosed with eating dis-
orders. Also, given the correlational nature of the data, we cannot make claims about
the temporal or causal relationships between the motivations and eating behaviors. Fi-
nally, we used only self-report data to measure the motivations and eating behaviors.
To begin to address these limitations, future research should collect psychometric
data on the entire 20-item measure using broader, more representative samples (e.g.,
across a diﬀerent range of ages and cultural backgrounds). The subscales should also
be related to clinically-accepted indices of eating disorders, and administered to clin-
ical samples. Longitudinal designs examining the development of motivations to eat
and their causes and consequences would shed light on the temporal aspects we
could not address with the present data. Experimental designs would help us under-
stand the causal mechanisms involved with these motivations (e.g., can the priming
of individual motivations cause diﬀerent types of eating behaviors?). Finally, as in
other research on motivations, non-self-report measures (e.g., projective measures,
behavioral observation) could be developed to assess motivations to eat.
In sum, this investigation extends knowledge about the psychological bases of eat-
ing behaviors, as prior research focused on only coping and compliance motivations
to eat. We do not intend this as necessarily a comprehensive model of all possible
reasons people eat, but of major psychological motivations for eating. As such, we
believe that in making parallels with motivational models in other health domains,
the Motivations to Eat measure is an important contribution.
This research was supported by Grant #AA08047 from the National Institute of
Alcohol Abuse and Alcoholism to M. Lynne Cooper. Thanks to Eileen Zurbriggen
for helpful comments regarding earlier drafts of this article.
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury
316 B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318
Allison, D. (1994). Handbook of assessment methods for eating behaviors and weight-related problems.
Thousand Oaks, CA: Sage.
Andersen, A. E. (1999). Gender-related aspects of eating disorders: A guide to practice. Journal of Gender-
Speciﬁc Medicine, 2, 47–54.
Arnow, B., Kenardy, J., & Agras, W. S. (1995). The Emotional Eating Scale: The development of a
measure to assess coping with negative aﬀect by eating. International Journal of Eating Disorders, 18,
Bentler, P. (1995). EQS Structural Equations Program manual. Encino, CA: Multivariate Software, Inc.
Bentler, P. M., & Bonnett, D. G. (1980). Signiﬁcance tests and goodness-of-ﬁt in the analysis of covariance
structures. Psychological Bulletin, 88, 588–600.
Bollen, K. A. (1989). Structural equations with latent variables. New York, NY: Wiley.
Bollen, K. A., & Long, J. S. (1993). Testing structural equation models. Newbury Park, CA: Sage.
Bordo, S. (1993). Unbearable weight: Feminism, Western culture, and the body. Berkeley: University of
Brisman, J., & Seigel, M. C. (1984). Bulimia and alcoholism: Two sides of the same coin. Journal of
Substance Abuse Treatment, 1, 113–118.
Brumberg, J. J. (2000). Fasting girls: The history of anorexia nervosa. New York: Vintage Books.
Conner, M., Norman, P., & Bell, R. (2002). The theory of planned behavior and healthy eating. Health
Psychology, 21, 194–201.
Cooper, M. L., Agocha, V. B., & Sheldon, M. S. (2000). A motivational perspective on risky behaviors:
The role of personality and aﬀect regulatory processes. Journal of Personality, 68, 1059–1088.
Cooper, M. L. (1994). Motivations for alcohol use among adolescents: Development and validation of a
four-factor model. Psychological Assessment, 6, 117–128.
Cooper, M. L., Frone, M. R., Russell, M., & Mudar, P. (1995). Drinking to regulate positive and negative
emotions: A motivational model of alcohol use. Journal of Personality and Social Psychology, 69, 990–
Cooper, M. L., & Shapiro, C. M. (1997). Motivations for health behaviors among adolescents. In J. A.
McNamara (Ed.), Creating the compliant patient (pp. 25–46). Ann Arbor, MI: Center for Human
Growth and Development, University of Michigan.
Cooper, M. L., Shapiro, C. M., & Powers, A. M. (1998). Motivations for sex and sexual behavior among
adolescents and young adults: A functional perspective. Journal of Personality and Social Psychology,
75, 1528–1558, Lead article.
Cox, M., & Klinger, E. (1988). A motivational model of alcohol use. Journal of Abnormal Psychology, 97,
Cox, M., & Klinger, E. (1990). Incentive motivation, aﬀective change, and alcohol use: A model. In M.
Cox (Ed.), Why people drink. New York: Gardner Press.
Floyd, F. J., & Widaman, K. F. (1995). Factor analysis in the development and reﬁnement of clinical
assessment instruments. Psychological Assessment, 7, 286–299.
Fredrickson, B. L., Roberts, T.-A., Noll, S. M., Quinn, D. M., & Twenge, J. M. (1998). That swimsuit
becomes you: Sex diﬀerences in self-objectiﬁcation, restrained eating, and math performance. Journal
of Personality and Social Psychology, 75, 269–284.
Gangley, R. M. (1988). Emotional eating and how it relates to dietary restraint, disinhibition, and
perceived hunger. International Journal of Eating Disorders, 7, 635–647.
Heatherton, T. F., & Baumeister, R. F. (1991). Binge eating as an escape from self-awareness.
Psychological Bulletin, 110, 86–108.
Horn, J. L., McArdle, J., & Mason, R. (1983). When is invariance not invariant: A practical scientistÕs
look at the ethereal concept of factor invariance. The Southern Psychologist, 1, 179–188.
Hu, L., & Bentler, P. M. (1999). Cutoﬀ criteria for ﬁt indexes in covariance structure analysis:
Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55.
Leary, M. R., Tchividjian, L. R., & Kraxberger, B. E. (1994). Self-presentation can be hazardous to your
health: Impression management and health risk. Health Psychology, 13, 461–470.
Leham, A. K., & Rodin, J. (1989). Styles of self-nurturance and disordered eating. Journal of Consulting
and Clinical Psychology, 57, 117–122.
B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318 317
Loehlin, J. C. (1998). Latent variable models: An introduction to factor, path, and structural analysis (3rd
ed.). Mahwah, NJ: Erlbaum Associates.
Macht, M., & Simons, G. (2000). Emotions and eating in everyday life. Appetite, 35, 65–71.
Marsh, H. W., & Hocevar, D. (1985). Application of conﬁrmatory factor analysis to the study of self-
concept: First- and higher-order factor models and their invariance across groups. Psychological
Bulletin, 97, 562–582.
Mintz, L. B., & Betz, N. E. (1998). Prevalence and correlates of eating disordered behaviors among
undergraduate women. Journal of Counseling Psychology, 35, 463–471.
Mintz, L. B., & Wright, D. (1993). Women and their bodies: Eating disorders and addictions. In E. Cook
(Ed.), Women, relationships, and power: Implications for counseling (pp. 211–246). Alexandria, VA:
American Counseling Association.
Mulaik, S. A., James, L. R., Van Alstine, J., Bennett, N., Lenni, S., & Stillwell, C. D. (1989). Evaluation of
goodness-of-ﬁt indices for structural equation models. Psychological Bulletin, 105, 430–445.
Neter, J., Kutner, M. H., Nachtsheim, C. J., & Wasserman, W. (1996). Applied statistical models (4th ed.).
Noll, S. M., & Fredrickson, B. L. (1998). A mediational model linking self-objectiﬁcation, body shame,
and disordered eating. Psychology of Women Quarterly, 22, 623–636.
Pliner, P., & Chaiken, S. (1990). Eating, social motivations, and self-presentation in women and men.
Journal of Experimental Social Psychology, 26, 240–254.
Roth, D. A., Herman, C. P., Polivy, J., & Pliner, P. (2001). Self-presentational conﬂict in social eating
situations: A normative perspective. Appetite, 36, 165–171.
Roth, G. (1993). Breaking free from compulsive eating. New York: Plume Books.
Sherwood, N. E., Crowther, J. H., Wills, L., & Ben-Porath, Y. S. (2000). The perceived function of eating
for bulimic, subclinical bulimic, and non-eating disordered women. Behavior Therapy, 31, 777–793.
Snook, S. C., & Gorsuch, R. L. (1989). Component analysis versus common factor analysis: A Monte
Carlo Study. Psychological Bulletin, 106, 148–154.
Snyder, M., & Cantor, N. (1998). Understanding personality and social behavior: A functionalist strategy.
In D. T. Gilbert & S. T. Fiske (Eds.), The handbook of social psychology (Vol. 2, 4th ed., pp. 635–679).
New York: McGraw-Hill.
Steptoe, A., Pollard, T. M., & Wardle, J. (1995). Development of a measure of the motives underlying the
selection of food: The food choice questionnaire. Appetite, 25, 267–284.
Stice, E., & Agras, W. S. (1999). Subtyping bulimic women along dietary restraint and negative aﬀective
dimensions. Journal of Consulting and Clinical Psychology, 67, 460–469.
Strahan, R., & Gerbasi, K. C. (1972). Short, homogenous versions of the Marlow-Crowne Social
Desirability Scale. Journal of Clinical Psychology, 28, 191–193.
Strauman, T. J., Vookles, J, Berenstein, V., Chaiken, S., & Higgins, E. T. (1991). Self-discrepancies and
vulnerability to body dissatisfaction and disordered eating. Journal of Personality and Social
Psychology, 61, 946–956.
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.
Stunkard, A. J., & Messick, S. (1988). The Eating Inventory. San Antonio, TX: Psychological Corporation.
van Strien, T., Frijters, J. E. R., Bergers, G. P. A., & Defares, P. B. (1986). The Dutch Eating Behavior
Questionnaire (DEBQ) for Assessment of Restrained, Emotional, and External Eating. International
Journal of Eating Disorders, 5, 295–315.
Waters, A., Hill, A., & Waller, G. (2001). Internal and external antecedents of binge eating episodes in a
group of women with bulimia nervosa. International Journal of Eating Disorders, 29, 17–22.
Widaman, K. F., & Reise, S. P. (1997). Exploring the measurement in variance of psychological
instruments: applications in the substance abuse domain. In K. J. Bryant, M. T. Windle, & S. G. West
(Eds.), The science of prevention: methodological advances from alcohol and substance abuse research
(pp. 281–324). Washington, DC: American Psychological Association.
Wolf, N. (1991). The beauty myth: How images of beauty are used against women. New York: W. Morrow.
318 B. Jackson et al. / Journal of Research in Personality 37 (2003) 297–318