Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Journal of Abnormal Psychology

2007, Vol. 116, No. 1, 188 197

Copyright 2007 by the American Psychological Association


0021-843X/07/$12.00 DOI: 10.1037/0021-843X.116.1.188

Thinness and Eating Expectancies Predict Subsequent Binge-Eating and


Purging Behavior Among Adolescent Girls
Gregory T. Smith

Jean R. Simmons

Kate Flory

Agnes M. Annus and Kelly K. Hill

University of Kentucky

Cleveland Clinic

University of South Carolina

University of Kentucky

Ones expectancies for reinforcement from eating or from thinness are thought to represent summaries
of ones eating-related learning history and to thus influence the development of binge-eating and
purging behavior. In a 3-year longitudinal study, the authors tested this hypothesis and the hypothesis that
binge eating also influences subsequent expectancy development. The authors used trajectory analysis to
identify groups of middle school girls who followed different trajectories of binge eating, purging, eating
expectancies, and thinness expectancies. Initial eating and thinness reinforcement expectancies identified
girls whose binge eating and purging increased during middle school, and expectancies differentiated
girls who began these problem behaviors from girls who did not. Initial binge-eating scores differentiated
among eating expectancy developmental trajectories. The onset of most behaviors can be understood in
terms of learned expectancies for reinforcement from these behaviors. The same model can be applied
to the risk for eating disorders.
Keywords: binge eating, purging, expectancies, risk factors

Individual women vary in the extent to which they have learned


to expect overgeneralized life improvement from thinness, and
they also vary in the extent to which they have learned to expect
eating to alleviate negative mood states (Hohlstein, Smith, &
Atlas, 1998). Our primary aim in this study was to demonstrate
that, among middle school girls, these expectancies predict the
subsequent onset of, and increase in, eating-disorder behaviors.
Our doing so is important for several reasons. First, the onset of
eating-disorder symptomlike behaviors appears to occur in early
adolescence (Gardner, Stark, Friedman, & Jackson, 2000; McKnight Investigators, 2003; Wertheim, Koerner, & Paxton, 2001).
Second, researchers are currently limited in their ability to predict
symptom onset, so there is a need to expand understanding of the
risk process (Stice, 2002). Third, as we describe below, eatingdisorder expectancy theory can integrate much of the existing risk
literature and extend that literature in useful ways (Fischer, Smith,
& Cyders, 2005).

quodale and Meehl (1953), and Bolles (1972). The theory postulates that one forms expectancies for the consequences of various
behaviors as a result of ones learning history. These expectancies
then influence ones future behavioral choices: One tends to
choose behaviors from which one expects rewards and avoid
behaviors for which one expects punishment. In other words, ones
expectancies for the consequences of a given behavior are a
summary of ones learning history and are thus the cognitive
mechanisms by which prior learning leads to subsequent behavior.
The application of this perspective by researchers has led to useful
examinations of many psychological phenomena, including psychopathology (Alloy & Tabachnik, 1984), affect (Carver &
Scheier, 1990), gambling behavior (Walters & Contri, 1998), and
risk for alcohol abuse (Smith, Goldman, Greenbaum, & Christiansen, 1995).

Expectancy Theory Applied to Eating Disorders

Expectancy Theory

Expectancy theory can be readily extended to eating disorders:


Because individuals are exposed to different learning experiences
concerning eating, dieting, and thinness, individuals form different
eating, dieting, and thinness expectancies. To the degree that some
women have come to associate eating with powerful reinforcers,
these women hold unusually strong expectancies for reinforcement
from eating and so pursue food with greater vigor. To the degree
that some women have come to associate thinness with powerful,
perhaps overgeneralized, reinforcers, these women hold strong
expectancies for reinforcement from thinness and hence pursue
thinness more strongly than do others. Extremes in eating and
dieting, such as eating-disorder symptoms, are thought to stem

Expectancy learning theory was first articulated by Tolman


(1932) and was further developed by Rotter (1954), MacCor-

Gregory T. Smith and Agnes M. Annus, Department of Psychology,


University of Kentucky; Jean R. Simmons, Department of Psychiatry and
Psychology, Cleveland Clinic, Cleveland, Ohio; Kate Flory, Department of
Psychology, University of South Carolina; Kelly K. Hill, Department of
Psychiatry, University of Kentucky.
Correspondence concerning this article should be addressed to Gregory
T. Smith, Department of Psychology, University of Kentucky, Lexington,
KY 40506-0044. E-mail: gsmith@uky.edu
188

PREDICTION OF BINGE EATING AND PURGING

from extreme or unusual learning histories (Fischer et al., 2005;


Hohlstein et al., 1998).
Eating-disorder expectancy theory may help researchers organize and integrate the existing risk literature. One set of documented risk factors can be understood as precursors of expectancies. Both the perception of pressure to be thin (McKnight
Investigators, 2003; Stice, 2001) and the experience of being
teased about ones weight (Gardner et al., 2000; Thompson, Coovert, Richards, Johnson, & Cattarin, 1995) longitudinally predict
symptom endorsement. They also predict increased dieting and
body dissatisfaction (Stice & Shaw, 2002; Stice & Whitenton,
2002; Wertheim et al., 2001), both of which predict subsequent
symptom endorsement (Ball & Lee, 2002; Cooley & Toray, 2001;
Killen et al., 1994). The studies of these factors have been important in our understanding of the risk process.
According to expectancy theory, these factors likely contribute
to expectancy formation. For example, if one endorses the perceived pressure item, Ive noticed a strong message from my
friends to have a thin body (Stice, 2001) or if one reports being
teased about ones weight, one is reporting experiences that contribute to the expectancy that dieting and thinness lead to reinforcement. Related to this, many items on the risk scale measuring
endorsement of the thin ideal (Stice, 2001) appear to represent the
specific expectancy that thinness increases attractiveness (e.g.,
Slender women are more attractive).
There are advantages to measuring overall expectancies. First,
because expectancies can be understood as the summary of what
individuals have learned, they are likely to reflect both those
influences that are represented in prior studies and other learning
influences not yet studied. Second, expectancies are more proximal to symptom expression in the risk process. Instead of measuring exposure to individual events that may influence ones
learning (such as pressure to be thin or being teased about ones
weight), one measures what one has actually learned, that is, one
measures the reinforcement one expects from thinness and dieting
(or from eating). Because individuals act to attain expected reinforcement (Bolles, 1972; MacCorquodale & Meehl, 1953), one is,
presumably, measuring the actual causal factor that results from
various learning experiences.
Hohlstein et al. (1998) found that expectancies for negative
reinforcement from eating (such as alleviation of negative mood)
and expectancies for overgeneralized life improvement from thinness (such as being more attractive, better respected, more selfreliant, less stressed, more able to conquer things, more capable,
and better able to cope with failures) differentiated anorexia nervosa patients, bulimia nervosa patients, and psychiatric controls
with 94% accuracy. These expectancies have been shown to crosssectionally correlate with symptom level in both adolescent and
adult samples (Hohlstein et al., 1998; MacBrayer, Smith, McCarthy, Demos, & Simmons, 2001; Simmons, Smith, & Hill, 2002;
Stice & Whitenton, 2002).
If expectancies do represent summaries of individuals learning
histories, it may be the case that symptom-expectancy influences
are reciprocal. Perhaps experiences of binge eating or purging
influence subsequent expectancies. This possibility has not been
investigated, although reciprocal symptom-expectancy influences
have been documented for adolescent drinking behavior (Smith et
al., 1995).

189

A Prospective Test of Eating Disorder Expectancy Theory


Our primary aim in this study was to test whether expectancies
predicted increases in eating disorder behaviors over time among
early adolescent girls. We proceeded in three steps. First, we
considered whether there were subgroups of middle school girls
that manifest different developmental trajectories in their reports
of binge-eating behaviors and purging behaviors. We anticipated
that some girls would consistently report few such behaviors, other
girls would have begun engaging in these behaviors prior to the
start of the study (Field, Camargo, Taylor, Berkey, & Colditz,
1999; McKnight Investigators, 2003; Wonderlich, Connolly, &
Stice, 2004) and so would consistently report many such behaviors, and other girls would begin to report these behaviors during
the middle school years. We therefore conducted trajectory analyses to test whether groupings of this kind fairly summarize
middle school girls reports.
Second, we tested whether expectancy predicts increases in the
target behaviors during the formative middle school years. Consistent with existing theory (e.g., Stice, 2001), we expected (a)
thinness expectancies to predict increases in both binge-eating and
purging behavior because the pursuit of thinness leads to binge
eating as dieting attempts fail and because of the belief that
purging facilitates thinness; and (b) the expectancy that eating
helps alleviate negative affect to predict increases in binge-eating
behavior for the following reasons. Neuroticism appears to be a
risk factor for eating disorders (Stice, 2002). For example, daily
diary studies indicate that women tend to binge eat on days when
they experience anxiety and stress (Crowther, Sanftner, Bonifazi,
& Shepherd, 2001). Further, laboratory studies show that the
induction of negative affect tends to lead to eating (Agras & Telch,
1998; Sanftner & Crowther, 1998). And, motivations to cope with
distress are positively associated with food consumption (Jackson,
Cooper, Mintz, & Albino, 2003).
We approached the measurement of binge eating and purging
with the assumption that endorsement of these problem behaviors
exists on a continuum in the population. There is considerable
evidence supporting this assumption (cf. Franko & Omori, 1999;
Stice, Ziemba, Margolis, & Flick, 1996). We therefore emphasized
the increase in bulimic-symptom expression rather than the onset
because we are not referring to the onset of the diagnosable
symptoms of the Diagnostic and Statistical Manual of Mental
Disorders (4th ed.; DSMIV American Psychiatric Association,
1994). Of course, reports of purging or binge-eating behavior by
middle school girls are of clinical concern, even when diagnostic
criteria for DSMIV disorders are not met.
Third, we tested whether there were different trajectories of
expectancy development over time. We conducted trajectory analyses on the development of the expectancies that eating helps
alleviate negative affect and on the expectancy that thinness leads
to overgeneralized life improvement. Doing so enabled us to test
whether the target behaviors of binge eating and purging appeared
to influence expectancy development. Because of evidence that
women eat in response to negative affect (Agras & Telch, 1998;
Sanftner & Crowther, 1998) and some women describe alleviation
of subjective distress as a motivation to binge eat (Jackson et al.,
2003), we anticipated that Time 1 binge eating would predict
subsequent increases in the expectancy that eating helps alleviate
negative affect. We did not expect the target behavior of binge

190

SMITH, SIMMONS, FLORY, ANNUS, AND HILL

eating to predict changes in thinness expectancies, nor did we


expect direct relations between purging and either binge-eating or
thinness expectancies.

Method
Participants
The participants were 394 middle school girls. They were first assessed
in the fall of their 7th-grade year (Time 1), which was their 1st year in a
2-year middle school. A total of 397 girls were approached to participate,
and 3 refused. The participants were followed through the fall of their
9th-grade year (their 1st year of high school). We began with 394 participating 7th graders and were able to locate and test 343 of them (87%) 1
year later (Time 2). At 2 years after the initial testing (Time 3), we were
able to locate 283 of them (83% of the Year 2 sample). As described below,
we used multiple imputation procedures to estimate missing values, which
enabled us to report results from the full sample of 394 girls. The mean age
at study outset was 12.84 years. Participants were predominantly Caucasian (78.7%); 10.5% identified themselves as African American; and the
remaining participants identified themselves as Asian, Hispanic, American
Indian, Arab, or East Indian. They were from diverse economic backgrounds: 26% reported family incomes that were less than $25,000, 50%
reported incomes that were between $25,000 and $50,000, and 24% reported incomes that were greater than $50,000.

Measures
Eating Expectancy Inventory Scale 1 (EEI; Hohlstein et al., 1998). We
used EEI Scale 1: Eating Helps Manage Negative Affect (e.g., when I am
feeling depressed or upset, eating can help me take my mind off my
problems). It has been shown to be internally consistent and associated
with eating disorder symptomatology in several adolescent samples
(MacBrayer et al., 2001; Simmons et al., 2002).
Thinness and Restricting Expectancy Inventory (TREI; Hohlstein et al.,
1998). The TREI is a single scale measure that reflects overgeneralized
expectancies for life improvement from thinness (e.g., I would feel less
stressed, in general, if I were thin). It has also been shown to be internally
consistent and concurrently predictive of eating-disorder symptomatology
in several adolescent samples (MacBrayer et al., 2001; Simmons et al.,
2002).
Bulimia TestRevised (Thelen, Farmer, Wonderlich, & Smith, 1991).
The Bulimia TestRevised is a paper-and-pencil measure. In two independent studies, it had sensitivity, specificity, and negative predictive
values over .90, and positive predictive values over .70 with respect to
DSMIV criteria as diagnosed by interview (Thelen, Mintz, & Vander Wal,
1996; Welch, Thompson, & Hall, 1993). The scale has perhaps most often
been used to measure individual differences in symptom endorsement
along a continuum (e.g., Hohlstein et al., 1998; Simmons et al., 2002). It
has a four-factor structure that includes binge-eating and purging factors
among adolescents (Vincent, McCabe, & Ricciardelli, 1999). We used two
slightly modified scales from the measure to reflect these two criteria. Ten
items reflected binge eating. One example is, I would presently rate
myself a compulsive eater (one who engages in episodes of uncontrolled
eating). Internal consistency for the scale was ! " .88. Eleven items were
chosen to reflect purging behavior (e.g., how often do you intentionally
vomit after eating?). Internal consistency for this scale was ! " .83.
Responses to items were on a 5-point Likert scale and were coded so that
higher scores reflected higher levels of the target behavior. For example,
responses to the vomiting frequency question ranged from less than once
a month or never to two or more times a week.

Data analytic method. We first identified groups of girls reporting


different trajectories of bulimic symptom endorsement across the 3 years of
the study. To do so, we used finite mixture modeling (Nagin, 2005). When
using this method, one assumes that the target population can be accurately
described as a mixture of distinct groups defined by their developmental
trajectories. In brief, longitudinal data are used to identify the number of
groups that best fits the data and to describe the shape of the trajectory for
each group. One can then calculate the probability of each individual
belonging to each of the trajectory groups that make up the model.
Individuals can then be assigned to the group to which the probability of
their belonging is the highest.
Several fit indices are used to determine the optimal number of groups
and the validity of the grouping result. The Bayesian information criterion
(BIC) becomes increasingly less negative with improvements in the fit of
the group structure. That statistic is supplemented by additional statistics
and rules of thumb for their application: When the average probability of
group membership is greater than .70 for each group and when the odds of
correct classification (OCC) are greater than 5 to 1, the identified group
structure is thought to fit well (Nagin, 2005). One also avoids group
structures with extremely small group sizes, out of concern for the stability
of the structure (Nagin, 2005). This method has been used profitably in a
number of recent studies (cf. Chassin, Pitts, & Prost, 2002; Flory, Lynam,
Milich, Leukefeld, & Clayton, 2004) and has been described in depth in
Nagin (2005). We used analysis of variance and planned contrasts to test
whether eating and thinness expectancies differentiated among the trajectory groups in expected ways.
We then used the same analytic approach to examine possible trajectories of eating expectancy and thinness expectancy development, and we
again used analysis of variance to test whether initial binge-eating and
purging scores differentiated among expectancy trajectory groups in the
expected ways.

Results
Participant Attrition Analysis and Treatment of
Missing Data
According to school officials, 12% of students moved from the
school system between Times 1 and 2, and 13% moved between
Times 2 and 3. Our retention rates of 87% and 83% indicated that
we retained almost all students still in the school system (we lost
44 students at Time 3 because of high school administrative
difficulties; those difficulties were unrelated to student characteristics). We compared study dropouts with retained students on
demographic, expectancy, and symptom measures. The two groups
did not differ on any demographic variable, except that dropouts
were more likely to live with only one parent, #2(1, N " 394) "
23.18, p $ .001. The two groups did not differ on binge-eating,
purging, or expectancy scores. We concluded that scores were
missing at random.
We therefore used multiple imputation to estimate missing data.
This procedure is thoroughly described by Allison (2003); it has
been shown to produce relatively unbiased population parameter
estimates. We used SAS Version 9.1 (PROC MI) to conduct the
imputation procedure as described by Allison (2003), and final
parameter estimates were based on 10 imputed data sets.

Procedure

Subgroups of Girls: Different Developmental Trajectories


of Symptom Endorsement

Data collection. Students were tested annually over a 12 day period


either in their regular classrooms or in one central location.

We used Nagins (2005) procedure to determine (a) whether


individual differences in girls trajectories of binge eating and

191

PREDICTION OF BINGE EATING AND PURGING

Binge Eating Scores

Binge Eating Behavior


40
35
30
25
20
15
10
5
0

N=30
N=302
N=48
N=14

Wave 1

Wave 2

Wave 3

Purging Behavior
50
Purging Scores

purging could be accurately characterized in terms of subgroups


and (b) what the number and shapes of the binge-eating and
purging trajectory groups were. We separately developed trajectory groups for binge eating and purging. For both group models,
we assigned each participant to the group for which she had the
highest probability of belonging.
We conducted two trajectory analyses, each on three annual
time points, with SAS Version 9.1 PROC TRAJ (censored normal;
Jones, Nagin, & Roeder, 2001) to model the trajectories as a
function of a measurement wave. The highest order polynomial we
tested was quadratic. For each of the two analyses, we first
specified two groups and then tested a series of models in which
we increased the number of groups and used the BIC, the average
probability of group membership, the OCC, and the group sample
size to evaluate model fits (Nagin, 2005). The BIC value became
progressively less negative from the two-group solution through
the six-group solution. However, five- and six-group solutions did
not have higher OCC values, did include groups with very small
sample sizes, and did not involve groups with substantively different trajectories from those apparent in the four-group solutions.
The four-group solution for binge eating had average group membership probabilities from .78 to .98, and the lowest OCC value
was 8:1. The four-group solution for purging had average group
membership probabilities from .86 to .98, and the lowest OCC
value was greater than 11:1. We therefore adopted four-group
solutions for both binge-eating trajectories and purging trajectories. The resulting trajectories are displayed in Figure 1.
Binge-eating behavior. As shown in the Figure 1 top panel,
302 of the 394 girls essentially reported no binge-eating behavior
at each of the three annual data collections spanning the period
from 7th grade through 9th grade. A group of 48 girls reported
moderate levels of binge-eating behavior during each of the three
measurement waves. A small group of 14 girls (3.55% of the
sample) appeared to endorse high levels of binge-eating behavior
during 7th grade, with their reports appearing to drop closer to
moderate levels by 9th grade. The trajectory describing the final
group of 30 girls is one of binge eating that increased across the
three measurement waves, with the largest increase occurring at
Time 3. The rate of linear increase was statistically significant,
t(29) " 6.03, p $ .001. The magnitude of the change for this group
can be understood in terms of time-specific z scores. At Time 1,
the mean binge-eating score for that group was z " %.08. At Time
3, the mean score was z " 1.65. Thus, this groups scores rose
from about the mean of the group to the 95th percentile of the
group.
Binge-eating scores are not symptom counts but rather are sums
of 10 Likert-type items. The following information suggests what
scores on the binge-eating scale mean practically. A typical response by girls in the moderately scoring groups at Time 3 or the
high scoring group at Time 1 would be as follows. In response to
In the last 3 months, on the average, how often did you binge eat
(eat uncontrollably to the point of stuffing yourself)? the average
response would be 23 times a month compared with the asymptomatic answer of once a month or less (or never). This rate of
binge eating has been referred to as indicative of a partial syndrome disorder (McKnight Investigators, 2003).
Purging behavior. As shown in the bottom panel of Figure 1,
of the 394 girls, 323 consistently reported essentially no purging
behavior across the 2-year, three-wave period. A group of 37 girls

40

N=30
N=323
N=37
N=4

30
20
10
0
Wave 1

Wave 2

Wave 3

Figure 1. Figure 1 depicts results of trajectory analyses on reported


binge-eating behavior and purging behavior from 7th through 9th grades.
The top panel presents trajectories of binge-eating behavior, and the bottom
panel presents trajectories of purging behavior. On the x-axis, Waves 13
refer to annual data collections at Times 13, which occurred in the fall of
each year. Scores from 10 to 15 reflect a virtual absence of reports of the
target behavior (endorsement of the fully asymptomatic response on half to
all of the questions); scores of 30 reflect endorsement of moderate levels of
target behaviors (a 30 reflects, approximately, an average endorsement of
a 3 on these 5-point Likert scales).

reported moderate levels of engagement in purging behavior. A


small group of 4 girls (1% of the sample) reported high levels of
purging behavior. The final group of 30 girls is characterized by a
trajectory of purging that increased across the three measurement
waves. Again, the rate of linear increase was statistically significant, t(29) " 7.49, p $ .001. The magnitude of increase for this
group was pronounced: from a Time 1 z score of .63 to a Time 3
z score of 2.18.
Substantively, Time 3 purging scores of this increasing group
and the chronically high group can be characterized as follows. In
response to the question, How often do you intentionally vomit
after eating? the average response would be two to three times a
month, compared with two alternatives describing less frequent or
no vomiting.

Prediction of Bulimic Symptom Onset and Increase


by Expectancy
Binge-eating behavior. We first tested whether expectancies
for overgeneralized reinforcement from thinness (TREI) and expectancies that eating helps alleviate negative affect (EEI), measured at Time 1, predicted membership in the binge-eating trajectory group. Both expectancies did. When analyzed individually,

192

SMITH, SIMMONS, FLORY, ANNUS, AND HILL

analysis of variance results confirmed the association between


TREI scores and group membership, F(3, 390) " 13.14, p $ .001,
&2 " .09, and between EEI scores and group membership, F(3,
390) " 30.54, p $ .001, &2 " .19. When corrected for the
relationship between the two kinds of expectancies, each continued
to be significantly associated with trajectory group membership.
We followed those omnibus tests with tests of specific contrasts.
Of primary interest was whether expectancies could differentiate
the group that reported no binge eating across the three waves from
the group that started with virtually no binge eating but reported
increases in binge-eating behavior across time. The contrast tests
we used to compare these two groups on both expectancy scores
were significant. For the TREI, t(1, 390) " 3.37, p $ .001, rpb "
.19. For the EEI, t(1, 390) " 2.71, p $ .005, rpb " .15. We
conducted one other contrast. The two groups with higher bingeeating scores at Time 1 differed from the two largely non-binge
eating Time 1 groups on the TREI, t(1, 390) " 3.10, p $ .001,
rpb " .25, and the EEI, t(1, 390) " 6.76, p $ .001, rpb " .40.
Purging behavior. We next tested whether expectancies for
overgeneralized reinforcement from thinness predicted membership in purging trajectory group. Time 1 TREI scores were associated with group membership, F(3, 390) " 13.50, p $ .001, &2 "
.09. Contrast analyses indicated that Time 1 thinness expectancies
differentiated the purging-increase trajectory group from the nonpurging group, t(1, 390) " 2.51, p $ .01, rpb " .14. Time 1 TREI
scores also differentiated the two highest scoring groups from the
two lowest scoring groups, t(1, 390) " 3.92, p $ .001, rpb " .27.
We did not anticipate that the expectancy that eating helps
alleviate negative affect would predict the purging trajectory. Time
1 EEI scores did differentiate among the four trajectory groups,
F(3, 390) " 5.43, p $ .001, &2 " .04. However, contrast tests
indicated the overall effect was due to initial EEI differences
between the two originally symptomatic groups, on the one hand,
and the two originally asymptomatic groups, on the other hand. No
other contrast tests produced significant differences.1
We have provided additional basic information concerning all
four trajectory solutions in Table 1. The table provides Time 1
thinness and eating expectancy means for each of the binge-eating
and purging trajectory groups, as well as Time 1 binge-eating and
purging means for each of the thinness expectancy and eating
expectancy trajectory groups. We also examined the overlap between membership in binge-eating and purging trajectory groups.
To do so, we arranged the two trajectory groupings sequentially
(the group continually reporting no target behaviors, the target
behavior increase group, the initially moderate behavior level
group, and the initially highest behavior level group). Group
membership was strongly associated, #2(9, N " 394) " 177.26,
p $ .001, and 78.68% of cases fell on the diagonal. Of the girls,
282 (72%) were in both the consistently non-binge eating and
consistently nonpurging groups. Of the 30 girls in the purgingincrease group, 14 were in the initially moderate or high bingeeating groups, and 9 were in the binge-eating increase group. A
smaller percentage of the binge-eating increase group was engaged
in purging behavior: Of 30 girls, 9 were in the initially moderateor high-purging groups and 5 were in the purging-onset group; half
(15) were in the consistently nonpurging group. Only 20 of the 302
girls in the consistently non-binge eating group were in the
moderate- or high-purging groups, and only 25 of the 323 girls in

the consistently nonpurging group were in the moderate or high


binge-eating groups.
Supplemental analyses of bulimic symptom onset. Because the
trajectory groups differed in Time 1 levels of binge eating and
purging, the trajectory comparisons do not address onset of those
behaviors. We therefore supplemented those analyses by constructing two groups of girls: those reporting no binge eating at
Time 1 and those reporting no purging at Time 1. We then used
Cox proportional-hazards modeling to determine whether Time 1
expectancies predicted subsequent binge-eating onset among original non-binge eaters. Time 1 thinness expectancy scores predicted
subsequent binge-eating onset ( p $ .001, hazard ratio " 1.30, i.e.,
a 1-point increase in average thinness expectancy item endorsement meant a 30% higher rate of beginning binge eating). Eating
expectancies were similar ( p $ .05, hazard ratio " 1.27, i.e., a
1-point increase in average eating expectancy item endorsement
meant a 27% higher rate of beginning binge eating), although
eating expectancies did not add incremental validity to the prediction from thinness expectancies. We also predicted purging onset
among nonpurgers, and we found that Time 1 thinness expectancy
scores predicted subsequent purging ( p $ .001, hazard ratio "
1.11), as did eating expectancies ( p $ .05, hazard ratio " 1.08).
Again, eating expectancies did not add incremental validity to
prediction from thinness expectancies.

Different Developmental Trajectories of


Expectancy Endorsement
We again used Nagins (2005) procedure to evaluate (a) whether
individual differences in girls expectancy trajectories could be
accurately characterized in terms of subgroups and (b) what the
number and shapes of the thinness expectancy and eating expectancy trajectory groups were. We developed trajectory groups separately for the two types of expectancies. For both group models,
we assigned each participant to the group for which she had the
highest probability of belonging. The highest order polynomial we
tested was quadratic. For each of the two analyses, we first
specified two groups and we then tested a series of models in
which we increased the number of groups and used the BIC, the
1
Because one of the groups on which each of these analyses of variance
were conducted is small (n " 14 for binge eating, and n " 4 for purging),
there is an increased risk that statistical assumptions might be violated. We
tested both sets of the four trajectory groups for heterogeneity of variance
for both thinness and eating expectancies, and there was no significant
heterogeneity (all Levene tests were nonsignificant). We also reran the
analyses combining the two highest scoring groups. For both sets of
trajectory groups and for both thinness and eating expectancies, the omnibus F tests remained significant, the contrasts in which we compared the
increasing groups with the asymptomatic groups did not change, and a
contrast in which we compared the combined two high symptom groups
with the average of the two low symptom groups remained significant.
Reported binge eating and reported purging were both positively skewed
in this sample. Although linear model analyses are rather robust to violations of normality (Darlington, 1990), we reran each of the above analyses
after conducting square-root transformations of the binge-eating and purging variables. Exactly the same pattern of significant findings emerged, and
parameter estimates were quite similar. We have reported results from the
untransformed variables here.

193

PREDICTION OF BINGE EATING AND PURGING

Table 1
Means and Standard Deviations of the Groups
TE
Group

EE
SD

BE
SD

P
SD

SD

Symptom trajectory groups on initial expectancies


Binge eating
Group 1
Group 2
Group 3
Group 4
Purging
Group 1
Group 2
Group 3
Group 4

302
30
48
14

2.97
3.90
3.95
4.53

1.36
1.70
1.71
1.62

1.82
2.34
2.93
3.65

0.96
1.08
1.17
0.92

323
30
37
4

3.02
3.71
4.27
5.62

1.40
1.52
1.70
1.38

1.97
2.27
2.67
2.63

1.05
1.09
1.34
0.87

Expectancy trajectory group classes on initial symptoms


Thinness expectancies
Group 1
Group 2
Group 3
Group 4
Group 5
Eating expectancies
Group 1
Group 2
Group 3
Group 4
Group 5

59
125
85
102
23

13.43
14.06
14.50
16.10
23.17

5.71
6.21
5.60
6.00
11.03

14.71
14.22
16.74
16.29
24.08

5.06
3.65
5.79
5.68
12.16

232
82
36
35
9

13.26
18.94
15.00
16.13
24.78

4.99
8.41
5.31
5.31
8.77

15.15
18.16
15.67
15.87
17.78

4.87
8.85
5.87
4.61
4.68

Note. Means for the expectancy measures are item means and are based on a 17 scale. Means for the symptom
measures are total scores. TE " thinness expectancies; EE " eating expectancies; BE " binge eating; P "
purging.

average probability of group membership, the OCC, and the group


sample size to evaluate model fits (Nagin, 2005).
For thinness expectancies, the BIC value became progressively
less negative from the two-group solution through the five-group
solution. The five-group solution had average group membership
probabilities from .78 to .86, and the lowest OCC value was over
9:1. For eating expectancies, the five-group solution also had the
least negative BIC value and average group membership probabilities from .76 to .93, and the lowest OCC value was greater than
6:1. We therefore adopted five-group solutions for both types of
expectancies. The resulting trajectories are displayed in Figure 2.
Thinness expectancy trajectories. As shown in the top panel of
Figure 2, there was essentially one relatively small group with
generally low expectancy endorsement (n " 59): These girls
average responses to items measuring expectancies for overgeneralized life improvement from thinness were in the completely
disagree to mostly disagree range. There was one group (n " 102)
that consistently endorsed average levels of expectancies: Their
typical score of 4 reflects a neither agree nor disagree point along
the 17 point scale. There was also a small group (n " 23) that
consistently endorsed high levels of thinness expectancies: Their
average score of approximately 6 represents a mostly agree response to thinness reinforcement expectancy items. In addition to
these three groups, there were two groups reporting expectancy
changes over time. One large group described increased thinness
expectancy endorsement (n " 125): The average score for that

group moved 2 years later from mostly disagree to the midrange of


the scale. The other group (n " 85) reported an average decline in
thinness expectancy endorsement, from an average response near
the midpoint (mean item response of 3.43 on a 17-point scale) to
an average response between mostly disagree and completely
disagree.
Eating expectancy trajectories. As shown in the bottom panel
of Figure 2, similar trajectory groups emerged for the expectancy
that eating helps alleviate negative affect. The largest trajectory
group consistently disagreed with these expectancy items (n "
232); a smaller group endorsed responses close to the midpoint on
the items 7-point scale (n " 82); and a very small group consistently endorsed the expectancy items a bit more strongly (n " 9).
For this last group, average responses were closest to slightly
agree. There were, again, two groups reporting eating expectancy
change. One groups (n " 35) scores increased from disagreement
to much higher scores (near the midpoint). The other group (n "
36) reported declines from scores near the midpoint to scores
reflecting disagreement with the expectancy items.

Binge Eating and Purging as Predictors of Expectancy


Development Trajectories
Thinness expectancies. Time 1 purging scores differentiated
among the five trajectory groups, F(4, 389) " 16.11, p $ .001,
&2 " .14. We followed this overall test with a series of planned

194

SMITH, SIMMONS, FLORY, ANNUS, AND HILL

Thinness Expectancies

Expectancy Score

7
6

N=59
N=125
N=85
N=102
N=23

5
4
3
2
1
0
Wave 1

Wave 2

Wave 3

Eating Expectancies

Expectancy Score

6
5

N=232
N=82
N=36
N=35
N=9

4
3
2
1
0
Wave 1

Wave 2

Wave 3

Figure 2. Figure 2 depicts results of trajectory analyses of thinness


expectancies (top panel) and eating expectancies (bottom panel). On the
x-axis, Waves 13 refer to annual data collections at Times 13, which
occurred in the fall of each year. On the y-axis, scores are average item
responses for items falling on a 17 Likert scale and ranging from completely disagree to completely agree.

contrasts. Initial purging scores did not differentiate between the


continually low thinness expectancy group and the increasing
expectancy group. The scores did, however, predict membership in
the thinness expectancy increasing group versus the decreasing
trajectory group, t(1, 389) " 3.17, p $ .001, rpb " .26. The
decreasing expectancy group, which was of course higher in Time
1 thinness expectancies, reported higher purging levels than did the
increasing expectancy group. We also contrasted the two groups
with lower Time 1 expectancy scores (the continually low group
and the increasing group) with the three groups with higher Time
1 scores. Initial purging scores did predict that contrast, t(1, 389) "
7.09, p $ .001, rpb " .24. The groups with higher initial thinness
expectancy scores also had higher initial purging scores.
Time 1 binge-eating scores also differentiated among the five
trajectory groups, F(4, 389) " 12.03, p $ .001, &2 " .11. We
conducted the same set of contrasts, and initial binge eating predicted only the contrast in which we compared the two groups with
lower Time 1 expectancy scores with the three groups with higher
Time 1 expectancy scores, t(1, 389) " 5.79, p $ .001, rpb " .18.
Initial binge eating did not differentiate between the increasing
thinness expectancy group and the decreasing thinness expectancy
group.
Eating expectancies. Time 1 binge-eating scores differentiated
among the five eating expectancy trajectory groups, F(4, 389) "

19.05, p $ .001, &2 " .16. Initial binge eating also differentiated
between the eating expectancy increase group and the group with
continually low expectancy endorsement, t(1, 389) " 2.57, p $
.005, rpb " .18. The trajectory group that began at moderate levels
of expectancy endorsement and then declined began with lower
binge-eating scores than did the two groups who maintained stable
high or moderate eating expectancy endorsement, t(1, 389) " 4.62,
p $ .001, rpb " .17. In addition, the three trajectory groups that
began with higher eating expectancy scores were significantly
higher in initial binge eating than were the other two groups, t(1,
389) " 5.11, p $ .001, rpb " .32.
Time 1 purging scores also differentiated among the five eating
expectancy groups, F(4, 389) " 4.06, p $ .005. The only significant purging contrast was our comparison of the three initially
high expectancy groups with the two initially low expectancy
groups: The former had higher average purging scores, t(1, 389) "
1.87, p $ .05, rpb " .17.

Discussion
This report describes a test of eating disorder expectancy theory,
which involves applying basic learning expectancy theory to the
problem of eating disorders. The theory holds that individuals
differ in their expectancies for reinforcement from thinness and
dieting and from eating because they have been exposed to different learning histories concerning dieting and food. Learned expectancies are a basis for future behavioral choices; expectancies
are thought to be the mechanism by which prior learning influences behavior. Extreme eating and dieting behaviors, that is,
binge eating and purging, are thought to result from extreme
expectancies, themselves the product of extreme learning histories.
The first important finding of this study concerned the different
trajectories of binge eating and of purging among these adolescent
girls. Early adolescent girls binge-eating behavior and their purging behavior can be characterized by four subgroups with different
trajectories of the target behavior across the 2 years of middle
school and the 1st year of high school. By middle school, girls
already differ in their engagement in eating-disorder behaviors.
Most girls were not involved in either target behavior: 72% reported essentially no binge eating or purging across the three
waves of data collection. However, there were also small numbers
of girls reporting binge eating and purging from the beginning of
middle school. It seems clear that risk assessment should begin
prior to middle school. Of particular importance for the present
study, there was a subgroup of girls who reported significant
increases in binge eating or purging during the middle school
years.
Among the highest scoring girls, the average rates of binge
eating or purging (such as through self-induced vomiting) were
two or three times per month. Although other authors refer to these
rates as reflective of partial syndrome disorders (cf. McKnight
Investigators, 2003), our focus is on bulimic behaviors and their
emergence, rather than on diagnosable eating disorders. We emphasize levels of binge eating and purging that are of obvious
concern to parents or health care professionals. The numbers of
girls in the highest scoring binge-eating and purging groups, although low, are consistent with previous literature. We found
3.55% of girls reporting binge eating more than monthly and 1%
of girls reporting purging more than monthly at Time 1. Wonder-

PREDICTION OF BINGE EATING AND PURGING

lich et al. (2004) interviewed girls of similar age and found 3.8%
to have binge eaten more than once and 2% to have engaged in
some compensatory behavior. McKnight Investigators (2003) and
Field et al. (1999) described similar findings.
A key test of eating disorder expectancies as putative risk
factors for bulimic symptom development involves their ability to
differentiate between girls who consistently report an absence of
symptoms and those who report increased levels of symptomatic
behavior over time. These tests, for this longitudinal period, provided clear support for the theory. Both eating and thinness expectancies differentiated the group of non-binge eating girls from
the group who began with little binge eating and increased their
binge eating during the study period. In addition, when a group of
girls not engaging in binge eating was identified, both types of
expectancies predicted the rate at which the girls would begin to
binge eat.
Thinness expectancies also differentiated the nonpurging group
from the group that began with little purging behavior and that
increased their purging over time. And, in a group of initially
nonpurging girls, thinness expectancies predicted the rate at which
girls would begin to purge. It thus appears that expectancies
predate and predict subsequent bulimic symptoms as hypothesized.
In this study, we also examined different trajectories of expectancy development during these formative years. For both the
expectancy that eating helps alleviate negative affect and the
expectancy that thinness leads to overgeneralized life improvement, we found five different trajectories of expectancy development. For both types of expectancies, there were three trajectory
groups characterized by relatively consistent endorsement of expectancy items: one group with low scores, one with moderate
scores, and one with relatively higher scores. It is noteworthy that
a substantial proportion of 7th-grade girls, in their 1st year of
middle school, were already endorsing eating and dieting expectancies at nontrivial levels and that their endorsement rates remained consistent across these transition years. Also, for both
types of expectancies, there was one group of girls who reported
significant increases in endorsement and one group who reported
significant decreases in expectancy endorsement over time. In this
sample, 18% of girls fell in trajectory groups marked by eating
expectancy change over time, and over 53% fell in trajectory
groups marked by thinness expectancy change. Thus, these transition years do seem to be ones in which expectancy levels change
for a meaningful portion of adolescent girls.
We also considered the possibility that initial binge-eating and
purging behaviors themselves influenced the subsequent trajectories of expectancy development. As hypothesized, higher initial
binge-eating behavior was associated more with the group whose
eating expectancies started low and increased over time, compared
with the group who endorsed continually low expectancies. In
addition, lower 7th-grade binge-eating scores predicted membership in the group whose eating expectancies declined over time, as
compared with groups who endorsed continually moderate or high
levels of expectancies.
These findings are consistent with the possibility that binge
eating influences subsequent expectancy development. The direction of the effects is noteworthy. The binge eaters tended to grow
in the belief that eating helps alleviate negative affect, and low
levels of binge eating were associated with a decline in that belief.
There was no evidence of a corrective process in which binge

195

eating led to reduced expectancies that eating helps alleviate


negative affect. This effect is consistent with previous literature; it
may well be that binge eating distracts one from ones experience
of subjective distress.
Although there were other relationships between initial binge
eating or purging and expectancy trajectory group membership, the
relationships all reflected associations between Time 1 symptoms
and Time 1 expectancies. For example, initial purging scores
differentiated between the thinness expectancy trajectory groups
that differed in initial expectancy levels, but the scores did not
differentiate between groups that started with similar expectancy
levels that then diverged over time.
From an expectancy theory perspective, those who have conducted research on perceived pressure, awareness of social messages, and weight-related teasing have done the important work of
identifying the nature and sources of many social learning influences. Those who measure expectancies, then, take advantage of
and extend that work by asking women directly about the degree
to which they expect reinforcement from thinness and dieting and
from eating. The thinness and dieting expectancy item, I would
handle myself better in social situations if I were thin, is a direct
statement of a benefit the respondent expects to receive from being
thin. By measuring expectancies directly, one is moving one step
closer to symptom expression, moving from identifying salient
learning events to measuring what one has actually learned. In a
sense, one is measuring the conclusions individuals have formed
from their learning histories as to the benefits they expect to
receive from either thinness or eating. By measuring those conclusions, one captures learning history influences represented by
prior studies as well as influences not represented in those studies.
Eating disorder expectancy theory also helps integrate the understanding of risk for eating disorders with the understanding of
precipitants for behavior in general (whether adaptive or maladaptive). There is considerable evidence that individuals choose behaviors from which they expect reinforcement (Bolles, 1972; Rotter, 1954), and there is evidence that dysfunctional behavior can be
explained in the same way (Darkes & Goldman, 1993; Smith et al.,
1995). That binge eating and purging can be similarly explained
implies that the same processes of learning and expectancy formation that influence other behaviors also influence eatingdisorder behaviors. Those who conduct research in eating-disorder
risk can only benefit from relying on advances in understanding
behavioral precipitants in other domains.
This study also has important limitations. Although the developmental trajectory class solutions fit the data well, they were
based on relatively small sample sizes (particularly for the symptomatic classes) and, in the cases of binge eating and purging, were
derived from nonnormal data. Both of those considerations raise
concerns about the replicability of the trajectory classes; replication on larger samples and on samples with higher percentages of
disordered girls is clearly necessary. In addition, all risk and
bulimic symptom reporting was done by questionnaire and was not
supplemented by interview data. Although there is considerable
evidence for the validity of both the expectancy and the behavior
measures, face-to-face interviews provide opportunities for clarification of terms and, perhaps, more precise and specific
assessment.
It is also important to recognize that these data do not demonstrate a causal role for eating-disorder expectancies. Clearly, the

196

SMITH, SIMMONS, FLORY, ANNUS, AND HILL

underlying theory assigns them a causal role, and expectancies


appear causal in other domains (Darkes & Goldman, 1993). However, in this study, though it is prospective, we could not rule out
the possibility that expectancies are predictive only as artifacts of
other causes. Experimental manipulations have been shown to alter
expectancies (Fister & Smith, 2004); to support the causality of
expectancy theory, one must show that expectancy modification
produces changes in symptom level.
In sum, the present findings provide clear support for eatingdisorder expectancy theory. Whether expectancies operate causally, as they do in other domains, and what the role of expectancies
in comprehensive risk models is remain to be worked out.

References
Agras, W. S., & Telch, C. F. (1998). Effects of caloric deprivation and
negative affect on binge eating in obese binge eating disordered women.
Behavior Therapy, 29, 491503.
Allison, P. D. (2003). Missing data techniques for structural equation
modeling. Journal of Abnormal Psychology, 112, 545557.
Alloy, L. B., & Tabachnik, N. (1984). Assessment of covariation by
humans and animals: The joint influence of prior expectations and
current situational information. Psychological Review, 91, 112149.
American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author.
Ball, K., & Lee, C. (2002). Psychological stress, coping, and symptoms of
disordered eating in a community sample of young Australian women.
The International Journal of Eating Disorders, 31, 71 81.
Bolles, R. C. (1972). Reinforcement, expectancy, and learning. Psychological Review, 79, 394 409.
Carver, C. S., & Scheier, M. F. (1990). Origins and functions of positive
and negative affect: A control-process view. Psychological Review, 97,
19 35.
Chassin, L., Pitts, S. C., & Prost, J. (2002). Binge drinking trajectories from
adolescence to emerging adulthood in a high-risk sample: Predictors and
substance abuse outcomes. Journal of Consulting and Clinical Psychology, 70, 6778.
Cooley, E., & Toray, T. (2001). Body image and personality predictors of
eating disorder symptoms during the college years. The International
Journal of Eating Disorders, 30, 28 36.
Crowther, J. H., Sanftner, J., Bonifazi, D. Z., & Shepherd, K. L. (2001).
The role of daily hassles in binge eating. International Journal of Eating
Disorders, 29, 449 454.
Darkes, J., & Goldman, M. S. (1993). Expectancy challenge and drinking
reduction: Experimental evidence for a mediational process. Journal of
Consulting and Clinical Psychology, 61, 344 353.
Darlington, R. B. (1990). Regression and linear models. New York:
McGraw-Hill.
Field, A. E., Camargo, C. A., Taylor, C. B., Berkey, C. S., & Colditz, G. A.
(1999). Relation of peer and media influences to the development of
purging behaviors among preadolescent and adolescent girls. Archives of
Pediatric and Adolescent Medicine, 153, 1184 1189.
Fischer, S., Smith, G. T., & Cyders, M. A. (2005). Integrating personality
and environmental risk factors for bulimia nervosa. In F. Columbus
(Ed.), Recent advances in eating disorder research (pp. 85108). New
York: NOVA Science Publishers.
Fister, S. M., & Smith, G. T. (2004). Media effects on expectancies:
Exposure to realistic female images as protective factor. Psychology of
Addictive Behaviors, 18, 394 397.
Flory, K., Lynam, D., Milich, R., Leukefeld, C., & Clayton, R. (2004).
Early adolescent through young adult alcohol and marijuana use trajectories: Early predictors, young adult outcomes, and predictive utility.
Development and Psychopathology, 16, 193213.

Franko, D. L., & Omori, M. (1999). Subclinical eating disorders in adolescent women: A test of the continuity hypothesis and its psychological
correlates. Journal of Adolescence, 22, 389 396.
Gardner, R. M., Stark, K., Friedman, B. N., & Jackson, N. A. (2000).
Predictors of eating disorder scores in children ages 6 through 14: A
longitudinal study. Journal of Psychosomatic Research, 49, 199 205.
Hohlstein, L. A., Smith, G. T., & Atlas, J. G. (1998). An application of
expectancy theory to eating disorders: Development and validation of
measures of eating and dieting expectancies. Psychological Assessment,
10, 49 58.
Jackson, B., Cooper, M. L., Mintz, L., & Albino, A. (2003). Motivations to
eat: Scale development and validation. Journal of Research in Personality, 37, 297318.
Jones, B., Nagin, D. S., & Roeder, K. (2001). A SAS procedure based on
mixture models for estimating developmental trajectories. Sociological
Methods and Research, 29, 374 393.
Killen, J. D., Barr, C. B., Hayward, C., Wilson, D. M., Haydel, K. F.,
Hammer, L. D., et al. (1994). Pursuit of thinness and onset of eating
disorder symptoms in a community sample of adolescent girls: A threeyear prospective analysis. The International Journal of Eating Disorders, 16, 227238.
MacBrayer, E. K., Smith, G. T., McCarthy, D. M., Demos, S., & Simmons,
J. (2001). The role of family of origin food-related experiences in
bulimic symptomatology. The International Journal of Eating Disorders, 30, 149 160.
MacCorquodale, K., & Meehl, P. E. (1953). Preliminary suggestions as to
a formalization of expectancy theory. Psychological Review, 60, 55 63.
McKnight Investigators. (2003). Risk factors for the onset of eating disorders in adolescent girls: Results of the McKnight longitudinal risk
factor study. American Journal of Psychiatry, 160, 248 254.
Nagin, D. S. (2005). Group-based modeling of development. Cambridge,
MA: Harvard University Press.
Rotter, J. B. (1954). Social learning and clinical psychology. Englewood
Cliffs, NJ: Prentice Hall.
Sanftner, J. L., & Crowther, J. H. (1998). Variability in self-esteem, moods,
shame, and guilt in women who binge. International Journal of Eating
Disorders, 23, 391397.
Simmons, J. R., Smith, G. T., & Hill, K. K. (2002). Validation of eating
and dieting expectancy measures in two adolescent samples. The International Journal of Eating Disorders, 31, 461 473.
Smith, G. T., Goldman, M. S., Greenbaum, P. E., & Christiansen, B. A.
(1995). Expectancy for social facilitation from drinking: The divergent
paths of high-expectancy and low-expectancy adolescents. Journal of
Abnormal Psychology, 104, 32 40.
Stice, E. (2001). A prospective test of the dual-pathway model of bulimic
pathology: Mediating effects of dieting and negative affect. Journal of
Abnormal Psychology, 110, 124 135.
Stice, E. (2002). Risk and maintenance factors for eating pathology: A
meta-analytic review. Psychological Bulletin, 128, 825 848.
Stice, E., & Shaw, H. E. (2002). Role of body dissatisfaction in the onset
and maintenance of eating pathology. Journal of Psychosomatic Research, 53, 985993.
Stice, E., & Whitenton, K. (2002). Risk factors for body dissatisfaction in
adolescent girls: A longitudinal investigation. Developmental Psychology, 38, 669 678.
Stice, E., Ziemba, C., Margolis, J., & Flick, P. (1996). The dual pathway
model differentiates bulimics, subclinical bulimics, and controls: Testing
the continuity hypothesis. Behavior Therapy, 27, 531549.
Thelen, M. H., Farmer, J., Wonderlich, S., & Smith, M. (1991). A revision
of the Bulimia Test: The BULITR. Psychological Assessment, 3, 119
124.
Thelen, M. H., Mintz, L. B., & Vander Wal, J. S. (1996). The Bulimia
TestRevised: Validation with DSMIV criteria for bulimia nervosa.
Psychological Assessment, 8, 219 221.

PREDICTION OF BINGE EATING AND PURGING


Thompson, J. K., Coovert, M. D., Richards, K. J., Johnson, S., & Cattarin,
J. (1995). Development of body image, eating disturbance, and general
psychological functioning in female adolescents: Covariance structure
modeling and longitudinal investigations. International Journal of Eating Disorders, 18, 221236.
Tolman, E. C. (1932). Purposive behavior in animals and men. New York:
Century Company.
Vincent, M. A., McCabe, M. P., & Ricciardelli, L. A. (1999). Factorial
validity of the Bulimia Test-Revised in adolescent boys and girls.
Behaviour Research and Therapy, 37, 1129 1140.
Walters, G. D., & Contri, D. (1998). Outcome expectancies for gambling:
Empirical modeling of a memory network in federal prison inmates.
Journal of Gambling Studies, 14, 173191.
Welch, G. Thompson, L., & Hall, A. (1993). The BULITR: Its reliability

197

and clinical validity as a screening tool for DSMIIIR bulimia-nervosa


in a female tertiary education population. International Journal of Eating Disorders, 14, 95105.
Wertheim, E. H., Koerner, J., & Paxton, S. J. (2001). Longitudinal predictors of restrictive eating and bulimic tendencies in three different age
groups of adolescent girls. Journal of Youth and Adolescence, 30,
69 81.
Wonderlich, S. A., Connolly, K. M., & Stice, E. (2004). Impulsivity as a
risk factor for eating disorder behavior: Assessment implications with
adolescents. International Journal of Eating Disorders, 36, 172182.

Received February 4, 2005


Revision received July 23, 2006
Accepted July 26, 2006 !

You might also like