Health Psychology

2008, Vol. 27, No. 2(Suppl.), S109 S117

Copyright 2008 by the American Psychological Association

0278-6133/08/$12.00 DOI: 10.1037/0278-6133.27.2(Suppl.).S109

What Mediates the Relationship Between Family Meals and Adolescent

Health Issues?
Debra L. Franko

Douglas Thompson

Northeastern University

Maryland Medical Research Institute

Sandra G. Affenito

Bruce A. Barton

St. Joseph College

Maryland Medical Research Institute

Ruth H. Striegel-Moore
Wesleyan University
Objective: To determine whether the frequency of family meals in childhood is associated with positive
health outcomes in adolescence through the mediating links of increased family cohesion and positive
coping skills. Design: Data were obtained from the National Heart, Lung, and Blood Institute Growth and
Health Study (NGHS), a 10-year longitudinal study of 2,379 black and white girls assessed annually from
ages 9 19. The mediational analysis framework of H. C. Kraemer and colleagues (2001) was used to test
the hypothesis that the frequency of family meals in childhood (Study Years 1 and 3) would be related
to health outcomes (Study Year 10) through the mediating links of family cohesion and coping skills
(Study Years 7/8), after adjusting for baseline (Year 1) demographics as well as previous levels of the
outcome variables (Years 5/6). Main Outcome Measures: Several measures of adolescent health
variables were included as outcome measures. These included the Perceived Stress Scale, three Eating
Disorders Inventory subscales (drive for thinness, body dissatisfaction, and bulimia), number of days of
alcohol and tobacco consumption, and engaging in extreme weight control behaviors (e.g., self-induced
vomiting). Results: More frequent family meals in the first 3 study years predicted greater family
cohesion and problem- and emotion-focused coping in Years 7 and 8. Family cohesion mediated family
meals and risk of smoking in Year 10. Problem-focused coping mediated family meals and both stress
and disordered eating-related attitudes and behaviors in Year 10. Conclusion: Eating together as a family
during childhood may have multiple benefits in later years.
Keywords: family meals, adolescence, adolescent health, smoking, disordered eating, stress

Familial variables are known to influence child and adolescent

health concerns, including nutrition, physical activity, and substance use (Kiesner & Kerr, 2004; Resnick, Harris, & Blum, 1993).
Eating together as a family (family meals) may be one component of family life associated with positive health outcomes. More
specifically, investigators have found that frequent family meals
are associated with better nutrition and reduced risk of unhealthy
weight control behaviors, substance use, teenage sexual intercourse, and suicidal risk (Borowsky, Ireland, & Resnick, 2001;
Fulkerson et al., 2006; Kingon & OSullivan, 2001; NeumarkSztainer, Hannan, Story, Croll, & Perry, 2003; Tepper, 1999).
Taveras and colleagues (Taveras et al., 2005) reported that eating
dinner together as a family was related to overweight prevalence at
study entry but was not associated in longitudinal analyses of their
data. In a single-wave survey of 4,746 middle and high school
students (ages 1118), frequency of family meals was evaluated by
asking how many times over the previous 7 days did all or most of
the family eat a meal together. After controlling for factors including
family connectedness and demographics, girls reporting more frequent family meals exhibited reduced substance use (cigarettes, alcohol, and marijuana), higher grade point average, fewer depressive
symptoms, and decreased risk of a suicide attempt (Eisenberg, Olson,

Debra L. Franko, Department of Counseling and Applied Educational

Psychology, Northeastern University; Douglas Thompson and Bruce A. Barton, Maryland Medical Research Institute, Baltimore, Maryland; Sandra G.
Affenito, St. Joseph College; Ruth H. Striegel-Moore, Wesleyan University.
This research was supported by a grant from the National Heart, Lung, and
Blood Institute (NHLBI; HL/DK71122) and contract HC55023-26 and cooperative agreement U01-HL-48941-44. Participating NGHS centers included
the following: Childrens Medical Center, Cincinnati, Ohio: Stephen R.
Daniels, MD, (principal investigator) and John A. Morrison, PhD (coinvestigator); Westat, Inc., Rockville, Maryland: George B. Schreiber, ScD
(principal investigator) and Ruth Striegel-Moore, PhD (co-investigator); and
University of California, Berkeley, California: Zak I. Sabry, PhD (principal
investigator) and Patricia B. Crawford, DrPH, RD (co-investigator). Maryland
Medical Research Institute, Baltimore, Maryland (Bruce A. Barton, PhD,
principal investigator) served as the data coordinating center. Program office
was NHLBI: Eva Obarzanek, PhD, MPH, RD (project officer 1992-2007) and
Gerald H. Payne, MD (project officer 19851991).
We acknowledge with gratitude the long-term commitment of all NGHS
participants and their families who contributed to this study and the NGHS
study personnel for their dedication to the project.
Correspondence concerning this article should be addressed to Debra L.
Franko, Department of Counseling and Applied Educational Psychology,
Northeastern University, 203 Lake Hall, Boston, MA 02115-5000. E-mail:
d.franko@neu.edu
S109

S110

FRANKO, THOMPSON, AFFENITO, BARTON, AND STRIEGEL-MOORE

Neumark-Sztainer, Story, & Bearinger, 2004), as well as less extreme

weight control behaviors and chronic dieting (Neumark-Sztainer,
Wall, Story, & Fulkerson, 2004).
It should be noted that the literature to date has not addressed the
issue of directionality between family meals and positive outcomes. Although it is hypothesized that eating together as a family
results in beneficial health behaviors, an alternate possibility is that
families in which children are already doing well are more likely
to have meals together. The causal direction can only be definitively assessed in an experimental design; because this would be
difficult to do, the second best approach is to examine longitudinal
data using predictive models, recognizing that there may be other
(unmeasured) factors that might explain the positive outcomes.
The aim of this study was to examine the prospective relations
between family meals and health outcomes and, further, to explore
the mechanisms that might explain how family meals exert this
positive impact. Eisenberg et al. (2004) suggested a number of
possible mechanisms, including increased family communication,
time spent together, or parental modeling of coping skills. It is
possible that frequent family meals might influence later health
outcomes by promoting healthful skills and attitudes. For example,
family meals may lead to improved coping skills due to discussions of how to solve everyday problems. In turn, improved coping
skills may lead to positive outcomes, for example, healthier eating,
decreased depressive symptoms, and less substance use. Improved
skills and attitudes are a mediator that links family meals with
later health outcomes.
The literature on family communication, connectedness, and
cohesion appears to support this assertion, particularly in the areas
of mental health concerns (Compton, Thompson, & Kaslow, 2005;
Garrison et al., 1997) and problematic eating behaviors (e.g., binge
eating, restrictive dieting), including clinical eating disorders
(Cooper, Whelan, & Woolgar, 2004; Gardner, Stark, Friedman, &
Jackson, 2000; Rodriguez Martin, Novalbos Ruiz, Martinez Nieto,
Escobar Jimenez, & Castro de Haro, 2004). In a study of adolescent girls using unhealthy weight control practices (e.g., extreme
dieting, vomiting, laxative use), family connectedness, positive family
communication, parental supervision/monitoring, and maternal presence were found to be protective factors against these unhealthy
behaviors (Fonseca, Ireland, & Resnick, 2002). Rodriguez Martin et
al. (2004) found that family dysfunction was associated with higher

Figure 1.

scores on multiple subscales of the Eating Disorder Inventory (EDI;

Garner, Olmstead, & Polivy, 1983) in a group of adolescents with
anorexia nervosa and bulimia nervosa.
Capitalizing on longitudinal data collected in the 10-year National Heart, Lung, and Blood Institute Growth and Health Study
(NGHS) that assessed family meals in childhood and health outcomes in adolescence, we examined potential mediators that might
explain the mechanisms by which family meals exert their impact.
We tested the hypothesis that the frequency of family meals in
childhood would be associated with positive health outcomes in
adolescence through the mediating links of increased family cohesion (e.g., better communication and more time together) and
positive coping skills (see Figure 1).

Method
The NGHS was a 10-year longitudinal study of 2,379 black and
white girls who were 9 or 10 years old at study entry (National
Heart, Lung, and Blood Institute [NHLBI], 1992). NGHS was
specifically designed to investigate the development of overweight
and cardiovascular disease risk factors in black and white girls; no
other racial or ethnic minority groups were included by design, and
boys were not studied. Girls were recruited from three study sites:
University of California at Berkeley, University of Cincinnati/
Cincinnati Childrens Hospital Medical Center, and Westat, Inc./
Group Health in Rockville, Maryland. Girls were interviewed
annually between Study Year 1 and Study Year 10; girls mean
ages were 9.5, 10.5 11.5, 12.5, 13.5, 14.5, 15.5, 16.5, 17.5, and
18.6 years from Study Years 1 through 10, respectively. Retention
was high in Years 2 4 (96%, 94%, 91%), declined to 82% in Year
7, and increased to 89% in Year 10. Eligibility criteria to participate in NGHS included the following: (a) child and parents indicated that they were either white or black; (b) girls were within 2
weeks of age 9 or 10 years; (c) child gave assent; and (d) parents
gave consent and completed a household demographic sheet. Parental education was categorized as 4 or more years of college
versus less than 4 years of college. Education was chosen over
income as a proxy of socioeconomic status, because NGHS data
were collected in three regions differing in median household
income and also because previous research has shown that education is a better predictor of health-related outcomes than income

Time of measurement and hypothesized relations examined in the present study.

MEDIATORS OF FAMILY MEALS

(Ross & Mirowsky, 1999). Participants age was recorded as age

at last birthday.
Data from all participants (N 2,379) were included in the
current study. Most measures were given in alternating years
(Years 1, 3, 5, 7) to reduce subject burden.

Measures
As shown in Figure 1, the analysis was designed to test the
hypothesis that the frequency of family meals in childhood (Study
Years 1 and 3) would be related to health outcomes (Study Year
10) through the mediating links of family cohesion and coping
skills (Study Years 7/8), after adjusting for baseline (Year 1)
demographics as well as previous levels of the outcome variables
(Years 5/6). Measures collected during these specific study years
were used to capture the time series of events shown in Figure 1
and to permit adjustment for levels of dependent variables at
earlier time points. All mediators and outcomes used in this analysis were measured in alternating study years, some in evennumbered years and others in odd-numbered years; for this reason,
in some cases it was necessary to combine multiple years (2/3, 5/6,
and 7/8) to represent blocks of time. Specifically, the family meals
measure was based on 2 study years of data (Years 1 and 3, not
measured in Year 2) to indicate both the temporal stability and the
frequency of family meal patterns. The mediators were measured
in Year 7 (family cohesion) or 8 (coping), with the same measures
at Years 3 or 2 (respectively) used as adjustment variables. The
outcomes were all measured in Study Year 10, with the same
measures at Years 5 (EDI subscales) or 6 (all other outcomes) used
as adjustment variables.
Demographics (obtained at Study Year 1). Baseline demographics were race/ethnicity (non-Hispanic black or white, selfreported at study entry using US Census categories), parental
education (completed 4 years of college vs. less education), and
number of parents in the household (two parents vs. one parent in
the household).
Family meals (obtained at Study Years 13). In Study Years 1
and 3, girls were asked How often do you eat with your parent(s), with response options usually/always, sometimes, or
never or almost never. The measure of family meals completed
by the girls in the current study was validated by (a) comparison
with a measure of family meals completed by a parent during the
same time, and (b) comparison with other measures completed by
the girls indicating solitary eating behaviors (reasoning that the
frequency of family meals should be inversely related to frequency
of solitary eating). These comparisons were made for the same
years of the study as the years in which the measure of family
meals was obtained (Years 1 and 3). A composite score was
constructed across Years 1 and 3 for the parent measure and for
each of the measures of solitary eating. These composite scores
were then compared to the primary measure of family meals
completed by the participants at the same point in time (Study
Years 1 and 3).
The data indicated significant associations between the participant and parent measures of family meals. Specifically, the polychoric correlation coefficient (indicating the strength of association among ordinal variables) was 0.43 (SE 0.04). The MantelHanzel chi-square test of non-zero correlation among ordinal
measures also indicated significant correspondence, 2(1, N

S111

1279) 108.80, p .0001. Furthermore, there were significant

inverse correlations between the participant family meals question
and the indicators of solitary eating (polychoric correlations ranging from 0.08, SE 0.03, for I eat while I do my homework
to 0.31, SE 0.03, for I eat alone). Thus the assessment of
family meals completed by the participants was strongly associated with parental report of family meals and was inversely associated with participant solitary eating behaviors.
Mediators (Study Years 7/8, controlled for at Years 2/3). In
this study, three potential mediating variables were studied: family
cohesion, problem-focused coping, and emotion-focused coping.
To investigate family cohesion and coping skills as potential
mediating factors, we used cohesion measured in Year 7 and
coping measured in Year 8 (these measures were not collected in
the same study years). To measure the girls perceptions of family
cohesion, the Cohesion subscale from the Family Adaptability and
Cohesion Evaluation Scale (FACES) III (Olson, Portner, & Lavee,
1985) was used; questions were reworded slightly to be appropriate for children. Psychometric data presented by Olson et al.
(1985) indicated testretest reliability of 0.93 at a 5-week interval,
and evidence of discriminant validity has been shown by low
correlations between cohesion and adaptability (r .03) and
cohesion and social desirability (r .35). In the NGHS sample,
internal consistency, as measured by Cronbachs alpha, was 0.82
in Study Year 3 and 0.89 in Study Year 7.
Two secondary subscales from the Coping Strategies Inventory
(CSI; Tobin, Holroyd, & Reynolds, 1984) measured positive coping styles: problem-focused engagement and emotion-focused engagement. Tobin et al. (1984) reported that testretest reliability
over a 2-week period ranged from 0.67 to 0.83. Construct validity
has been shown in several studies examining the relationship
between the CSI and other coping and stress measures, and a
hierarchical factor analysis found support for the structure of the
CSI (Tobin, Holroyd, Reynolds, & Wigal, 1989). Cronbachs
alpha for each scale for Years 2 and 8 were as follows: problemfocused coping, 0.79 and 0.79; and emotion-focused coping, 0.84
and 0.90, respectively.
Outcomes (Study Year 10, controlled for at Years 5/6). Table
1 shows the outcome variables used in the present study and how
they were coded. Stress was measured using the Perceived Stress
Scale (PSS; Cohen, Kamarck, & Mermelstein, 1983), a 14-item
scale measuring the degree to which situations occurring in the
previous month are appraised as stressful. Psychometric data on
the PSS indicate that it is a reliable and valid measure. Testretest
reliability has been reported to be 0.85 over a 2-day interval
(Cohen et al., 1983; Hershberger, 1990). Evidence of concurrent and
predictive validity has been found between the PSS and depressive
symptoms (0.76) as well as the PSS and physical symptoms (0.64;
Cohen et al., 1983). In the NGHS sample, Cronbach alpha was 0.89
in Year 6 and 0.84 in Year 10.
Further, three subscales from the EDI (Garner et al., 1983) were
included as outcome variables: drive for thinness (DT; 7 items),
body dissatisfaction (BD; 9 items), and bulimia (7 items). Extensive data support the reliability and validity of this measure
(Franko et al., 2004; Garner et al., 1983; Wear & Pratz, 1987). In
the NGHS sample, Cronbach alphas for Years 5 and 10 were 0.88
and 0.92 for DT, 0.89 and 0.93 for BD, and 0.75 and 0.86 for
bulimia, respectively.

FRANKO, THOMPSON, AFFENITO, BARTON, AND STRIEGEL-MOORE

S112

Table 1
Definition/Coding and Descriptive Statistics for Measures Included in the Analysis (Demographics, Outcomes and Factors That May
Mediate the Relation Between Family Meals and Outcomes)
Measure
Demographics (measured at baseline)
Race/ethnicity
Parental education
Two-parent household
Mediators (measured in Years 7/8)
Family cohesion
Problem-focused engagement
Emotion-focused engagement
Outcomes (measured in Year 10)
Perceived stress
Extreme weight control behaviors
Drink alcohol
Smoke cigarettes
Smoke daily
Drive for thinness
Body dissatisfaction
Bulimia

Sample sizea

%b

White/black
4 Years of college education vs. less education (most
educated parent)
Two vs. one parent(s) in household

2,379
2,376

51.0% White/49%
35.1% 4 yrs/64.1%

2,378

FACES cohesion score (range 936)

Problem-focused engagement score (range 024)
Emotion-focused engagement score (range 024)

1,882
1,964
1,964

Perceived Stress Scale score (range 056)

Vomited, took diet pills, or took laxatives, ipecac, or
diuretics to control weight in past 30 days versus
engaged in none of those activities
Drank alcoholic beverages on 1 or more days in past 30
(yes/no)
Smoked cigarettes on 1 days in past 30 (yes/no)
Smoke cigarettes every day or nearly every day (yes/no)
EDI drive for thinness subscale score (range 021)
EDI body dissatisfaction subscale score (range 027)
EDI bulimia subscale score (range 021)

2,070
2,059

68.5% two parents/31.5%

Mean (SD)
22.4 (5.9)
14.5 (4.2)
15.1 (5.5)
Mean (SD) or %
22.8 (8.0)
6.8%

2,069

50.2%

2,067
2,069
2,070
2,066
2,071

33.2%
19.1%
5.1 (5.7)
9.1 (7.9)
1.0 (2.3)

Definition/coding

Sample sizes shown are prior to imputation of missing values via the hot-deck method. For all measures, the sample size postimputation was
2,379. b Descriptive statistics are for the National Heart, Lung, and Blood Institute Growth and Health Study (NGHS) sample prior to imputation of
missing values. For binary measures, the percentages postimputation were within 1% of the percentages prior to imputation. For continuous measures,
means and standard deviations postimputation were within 2% of the means and standard deviations prior to imputation.

In addition, participants were asked to report the number of days

in the previous 30 days when they had consumed alcohol, engaged
in extreme weight control behaviors, or smoked cigarettes.

Statistical Analysis
To test the hypothesis that family cohesion and coping skills
mediate the relation between family meals and outcomes, the analytic
framework of Kraemer and colleagues was used (Kraemer, Stice,
Kazdin, Offord, & Kupfer, 2001; Scott, Kraemer, Essex, & Kupfer,
2004). This framework may be viewed as a variant of the causal
steps approach to analysis of mediation (MacKinnon, Lockwood,
Hoffman, West, & Sheets, 2002). Mediation is inferred from a specific pattern of results across a series of regression models (the causal
steps) involving an outcome (O) and two possible risk factors (A and
B). A separate model is created for each outcome and pair of risk
factors. Although techniques such as path analysis or structural equation modeling would be possible alternative approaches, the framework of Kraemer et al. (2001) was chosen because it was specifically
designed to address mediation in longitudinal data and thus was an
excellent fit given the data and analytic objectives of the current study.
The framework defines mediation as follows. Assuming that A
and B are two possible risk factors for outcome O, B mediates the
association between A and O if and only if three conditions are
met: (1) A occurs before B; (2) A and B are significantly associated
with one another; and (3) in the regression model O A B
A B, either (a) there is a significant main effect of A as well as
a significant main effect of B or a significant A B interaction
(partial mediation), or (b) there is a significant main effect of B and
neither a significant main effect of A nor a significant A B
interaction (full mediation).

In the Part 1 models, mediators (B) were modeled as a function

of family meals (A) represented as a three-category ordinal variable (always/usually, sometimes, never or almost never). Conceptually, this is similar to a nonequivalent groups design (Cook &
Campbell, 1979) where frequency of family meals is analogous to
a treatment that is not randomly assigned. Because there was no
random assignment, the groups may differ in other ways besides
the frequency of family meals. Therefore, all models reported in
this article (Parts 1 and 2) were adjusted for study site, race,
parental education and one- versus two-parent household. In addition, the Part 1 models were adjusted for the mediator measured
in the same timeframe as family meals (i.e., Year 2 for coping and
Year 3 for family cohesion) and the interaction between the latter
and family meals, allowing us to assess prospective associations
between family meals and mediators after adjusting for prior
differences in the mediators.
In the second type of model (Part 2 models: O A B A
B), Year-10 outcomes served as the dependent variable (O) and the
independent variables of interest were family meals (A; Years 1
and 3), mediators (B; Years 7/8), and the interaction between the
two (A B). Each possible pair of mediators and outcomes was
examined in a separate model. In addition to the specific mediator
of interest in a given model, models were adjusted for the other
two mediators. Also, the outcome measured in Year 5/6 was
included in the model to estimate the association between a given
mediator and outcome (Year 10), holding constant the outcome
measured at a prior time. This allowed us to examine whether a
given mediator was associated with differences in the outcome
(Year 10) among girls who started out at the same level of the
outcome (Years 5/6).

MEDIATORS OF FAMILY MEALS

All Part 1 and Part 2 models were computed with PROC

GENMOD in SAS v9.1 (SAS Institute, Cary, NC) with appropriate
distribution and link functions depending on whether the outcome
was continuous or binary. We predicted that frequency of family
meals would be related to the mediators and the mediators would
in turn be predictive of each of the outcomes at Year 10. Therefore,
p .05 was used as criterion of statistical significance for these
main effects. Because no interactions were predicted, interactions
were judged at a more conservative 0.01 level to reduce the
possibility of Type I errors.
Effect sizes are expressed as r, the coefficient of partial correlation, representing the marginal, additional association with an
outcome attributable to a variable of interest in a model that
already takes into account the association due to a set of adjustment variables (Neter, Wasserman, & Kutner, 1990; for categorical outcomes, the coefficient was based on Nagelkerke, 1991).
Rules of thumb for categorizing r (small, medium, large) are
provided by Kraemer et al. (2003); for example, r 0.3 is
categorized as small.
Prior to analysis, all variables were center coded as described by
Kraemer and Blasey (2004); continuous variables are centered on
the median and binary variables are coded 0.5 versus 0.5, which
enables the results (e.g., model coefficients) to be interpreted as
estimates at average levels of the adjustment variables. As shown
in Tables 1 and 2, for the analytic variables, missing data ranged
from 0% (race) to 21% (family cohesion). Missing data were
imputed in SOLAS (v. 3.20, Statistical Solutions Ltd., Saugus,
MA) via hot decking. Briefly, each missing value was replaced by
a nonmissing value from a randomly selected girl with a similar
demographic profile; this method preserves the distributions of
analytic variables including the variance (Allison, 2002). Results
from analyses with missing data imputed via hot decking were
compared with results from parallel complete case analyses (i.e.,
excluding participants with missing data on any variable) to ensure
that the method of handling missing data did not have a major

S113

impact on the results. Unless noted otherwise, results reported in

the following section are based on data following imputation.

Results
Family Meals (Years 1 and 3)
Across Study Years 1 and 3, few girls (less than 7%) said that
they never or almost never ate with their parents in either year;
the modal pattern, accounting for slightly less than half of the girls,
was sometimes in both years or usually/always one year and
sometimes the other year. However, almost as many girls said
that they usually or always ate with their parents in both years.

Direct Effects of Family Meals on Outcomes

Family meals was measured at Years 1 and 3, and the main
outcome measures were measured at both Years 5/6 and Year 10.
The results of the direct effects analysis are shown in Table 2. The
means and percentages shown are adjusted for differences in study
site, race, parental education, and number of parents in the household. Table 2 indicates that the associations are generally stronger
when one looks at outcomes measured at Years 5/6 than at Year
10. There are direct effects at 5/6 and/or 10 for the following
variables: stress; smoking; and drive-for-thinness, body dissatisfaction, and bulimia subscales of the EDI.
Part 1 models: Association between family meals (Years 1 and
3) and hypothesized mediators (Years 7/8). The first set of analyses focused on the Part 1 models used to estimate the association
between family meals (A) and mediators (B). The mediators were
family cohesion, problem-focused coping, and emotion-focused
coping. As shown in Table 3, more frequent family meals (Study
Years 1 and 3) were predictive of greater family cohesion and
greater levels of problem- and emotion-focused coping (Years 7/8)
(all ps .0001). For all mediators, eating family meals some-

Table 2
Direct Effect of Family Meals on Outcome Variables: Mean (SD) or Percent of Outcome Measure by Time When the Outcome Was
Measured (Years 5/6 vs. 10) and Frequency of Family Meals
Frequency of family meals
(combined pattern, Years 1 and 3)
Outcome
Stress
Extreme weight control
Drink alcohol
Smoke cigarettes
Smoke daily
Drive for thinness
Body dissatisfaction
Bulimia

When outcome
was measured

Never or
almost never

Sometimes

Usually/always

P value of statistical test

(likelihood ratio 2 with 2 df)

Years 5/6
Year 10
Years 5/6
Year 10
Years 5/6
Year 10
Years 5/6
Year 10
Years 5/6
Year 10
Years 5/6
Year 10
Years 5/6
Year 10
Years 5/6
Year 10

23.56 (0.52)
23.36 (0.54)
4.2%
6.1%
19.9%
47.5%
13.5%
27.5%
4.1%
17.5%
5.18 (0.38)
4.79 (0.39)
8.30 (0.48)
8.73 (0.53)
2.14 (0.20)
1.05 (0.15)

24.35 (0.24)
23.20 (0.24)
4.1%
6.0%
21.6%
51.2%
13.1%
33.2%
3.7%
18.2%
4.71 (0.17)
5.23 (0.18)
7.52 (0.21)
9.32 (0.24)
1.64 (0.09)
0.94 (0.07)

23.41 (0.26)
22.08 (0.27)
3.9%
6.8%
20.9%
51.7%
9.9%
31.2%
3.1%
14.6%
4.18 (0.19)
5.06 (0.19)
7.02 (0.23)
8.87 (0.26)
1.38 (0.10)
0.92 (0.08)

.03
.007
.96
.73
.86
.58
.05
.27
.65
.09
.03
.53
.05
.35
.003
.77

S114

FRANKO, THOMPSON, AFFENITO, BARTON, AND STRIEGEL-MOORE

times versus never or almost never was associated with estimated mediator increases between 0.42 and 0.58 points (depending
on the specific mediator), whereas eating family meals usually/
always was associated with additional mediator increases of 0.26
to 0.38 points in the mediators. The models controlled for prior
assessments of the mediators; therefore, the mean differences
shown in Table 3 are interpreted as differences in the mediators
(e.g., family cohesion) at Years 7/8, with the corresponding mediator (e.g., family cohesion) held constant in prior years.
Part 2 models: Association between family meals (Years 1 and
3), mediators (Years 7/8), and outcomes (Year 10). The association of family meals with the mediators (Part 1 models) is a
necessary but not sufficient condition for mediation as defined
previously. For mediation, a specific pattern of results must also be
observed in the Part 2 models (outcome family meals mediator family meals x mediator).
Outcomes in Year 10 were modeled as a function of family
meals (Years 1 and 3), mediators measured during Years 7/8, and
interactions between the two. Because only the main effects and
interaction are relevant to the mediation hypothesis, only these are
described. Table 4 shows model coefficients representing the mediator main effects (linear slopes for continuous outcomes, odds
ratios for binary outcomes).
There were significant main effects of family cohesion in the
models of cigarette smoking and smoking daily. Specifically, higher
levels of family cohesion in Years 7/8 were related to a reduced
probability of any smoking in the previous 30 days as well as daily
smoking in Year 10. The effect sizes were in the small range (r 0.12
for daily smoking and 0.07 for any smoking). As an additional
indicator of effect size, a 10% increase in the family cohesion scale
predicted an 8.1% decrease in risk of smoking 1 or more days in the
last 30 days and a 12.8% decrease in risk of smoking every day. These
results from the Part 2 models, accompanied by the significant association between family meals and family cohesion (Part 1 models), are
consistent with the hypothesis that cohesion mediates the relation
between family meals and smoking. In the models involving cohesion, there were no significant main effects of family meals nor were
there any significant cohesion x family meals interactions. Cohesion
was not significantly associated with any of the other outcomes
measured in Year 10.
There was a significant main effect of problem-focused coping
in the models of stress and the EDI drive-for-thinness and bulimia

subscales. Greater problem-focused coping in Years 7/8 was associated with reduced stress and lower drive-for-thinness and
bulimia scores. All of the effect sizes were in the small range (r
0.08 to 0.14). As an additional indicator of effect size, a 10%
increase in the problem-focused coping scale was associated with
decreases of 1.1%, 0.9%, and 0.6% in stress, drive for thinness,
and bulimia, respectively. For smoking, there was also a significant interaction of family meals x problem-focused coping ( p
.008), due to the fact that there was a negative association of
problem-focused coping and smoking in the never or almost
never family meals group but no association in the other two
groups; however, there were no significant main effects of family
meals or problem-focused coping in this model. There were no
other main effects or interactions in the models relating problemfocused engagement to the outcomes. These results, together with
the significant association between family meals and problemfocused engagement (Part 1 models), are consistent with the hypothesis that problem-focused coping mediates the relation between family meals and these three outcomes (i.e., stress, EDI
drive-for-thinness score, and EDI bulimia score). However, the
presence of an interaction without either main effect suggests that
problem-focused coping does not mediate the relation between
family meals and smoking. Problem-focused coping was not significantly associated with any of the other outcomes measured in
Year 10.
Emotion-focused coping was not significantly associated with
any of the outcomes. Thus, the analysis did not reveal evidence
that emotion-focused coping mediates the relation between family
meals and any of the outcomes measured in Year 10.
Parallel complete case analyses (excluding from the analysis
participants with missing data for any variable) yielded results
very similar to the analyses with missing data imputed via hot
decking for most variables. The primary difference between the
two methods of handling missing data was that variance estimates
(standard errors, 95% confidence interval [CI]) tended to be
slightly greater in complete case analysis (which is to be expected
due to the reduced sample size in the latter), but this led to a
difference in conclusions about the effects of interest in only one
situation: For smoking, the estimated effect of family cohesion was
significant with hot decking (odds ratio [OR] 0.97, 95% CI
0.951.00, p .02) but not in complete case analysis (OR 0.97,
95% CI 0.94 1.00, p .08).

Table 3
Adjusted Mean (SE) of Mediators (Years 4 9) by Frequency of Family Meals (Years 1 and 3 Combined)a
Frequency of family meals
(combined pattern, Years 1 and 3) N 2379
Mediator (Years 49)

Never or
almost never

Sometimes

Usually/always

Statistical testb

Effect size
estimatec

Family cohesion (FACES)

Problem-focused coping
Emotion-focused coping

21.84 (0.39)
13.82 (0.28)
14.48 (0.37)

22.27 (0.17)
14.24 (0.13)
15.06 (0.17)

22.53 (0.19)
14.58 (0.14)
15.45 (0.18)

2(1) 101.02, p .0001

2(1) 24.51, p .0001
2(1) 33.91, p .0001

0.04
0.07
0.05

The table shows means by frequency of family meals, after adjustment for site, race, parental education, one- vs. two-parent household, the mediator
measured in Year 2 or 3, and the interaction between the latter and family meals. bLikelihood-ratio chi-square test of the main effect of family meals in
a model with the potential mediating factor as the dependent variable, adjusting for the variables listed in the previous note. c Effect size estimates are
expressed as r, the coefficient of partial correlation, interpreted as the marginal contribution of family meals to prediction of the corresponding mediator,
after accounting for the effects of race, study site, parental education, and number of parents in the household. One rule of thumb is that 0 r 0.30
indicates a small effect.

MEDIATORS OF FAMILY MEALS

S115

Table 4
Modeled Association Between Potential Mediating Factors and Outcomes
Mediator measured in Years 7/8
Outcomes (Year 10)

Family cohesion

Stress
Extreme dieting
Drink alcohol
Smoke cigarettes
Smoke daily
Drive for thinness
Body dissatisfaction
Bulimia

0.06 (0.03)
0.97 (0.931.00)
0.99 (0.971.00)
0.97 (0.951.00)*b
0.95 (0.930.98)***
0.02 (0.02)
0.05 (0.03)
0.01 (0.01)

Problem-focused
engagement

Emotion-focused
engagement

0.25 (0.05)S****a
0.99 (0.941.05)
1.01 (0.981.04)
0.97 (0.941.00)
0.97 (0.931.01)
0.08 (0.04)*
0.09 (0.05)
0.05 (0.01)***

0.01 (0.04)
1.01 (0.971.05)
1.01 (0.991.03)
1.01 (0.981.03)
1.02 (0.991.04)
0.01 (0.03)
0.03 (0.03)
0.02 (0.01)

Note. The table displays hypothesis test results (indicated by asterisks) and effect size indicators (beta coefficients and standard error for continuous
outcomes or odds ratios and 95% confidence interval for binary outcomes). FACES Family Adaptability and Cohesion Evaluation Scale.
a
Example of interpretation (continuous outcome): For each 1-point increase on the problem-focused engagement scale, girls stress scores are estimated
to decrease by 0.25 points, after adjustment for the other variables included in the analysis. bExample of interpretation (binary outcome): After adjustment
for the other variables included in the analysis, for each 1-point increase on the FACES scale, the estimated odds of smoking cigarettes is 0.97 times as
great; or equivalently, for each 1-point increase in FACES, the estimated odds of not smoking increases by 1/0.97 1.03.
Main effect of potential mediator: * p 05. ** p .01. *** p .001. **** p .0001.

Discussion
Family meals appear to be associated with a variety of childhood and adolescent positive health behaviors (Mamun, Lawlor,
OCallaghan, Williams, & Najman, 2005). This study sought to
increase our understanding of factors that mediate this relationship
by investigating how family meals might foster adolescent health.
Eating meals with parents was a common occurrence for girls ages
9 12, with fewer than 10% reporting that they never or almost
never ate with their parents during both study years. Furthermore,
girls who ate with their parents more often in Study Years 1 and 3
exhibited higher levels of family cohesion and problem- and
emotion-focused coping in Years 7/8, even after adjusting for these
mediators measured concurrently with family meals (Years 2/3).
This finding suggests that frequent family meals foster closer
family ties and may be helpful in teaching children coping skills.
Our data are consistent with Fulkerson, Neumark-Sztainer, and
Story (2006), who found associations between a positive atmosphere and frequency of family meals, suggesting that family
meals may enhance family communication and provide good role
modeling.
Girls who exhibited greater family cohesion (Years 7/8) had a
lower risk of smoking one or more times in the previous 30 days
as well as daily smoking (Year 10), even after adjusting for
frequency of smoking in earlier years (Year 5/6). Along with the
significant association between family meals and family cohesion,
this is consistent with the hypothesis that family cohesion mediates
family meals (Years 1 and 3) and risk of occasional and daily
smoking in Year 10. In addition, girls who exhibited higher levels
of problem-focused coping in Years 7/8 exhibited lower levels of
stress, drive for thinness, and bulimic symptoms in Year 10; the
evidence suggests that problem-focused coping mediates the relation between family meals and variables measuring stress, drivefor-thinness, and bulimia scores. Taken together, these findings are
consistent with the possibility that family meals positively impact
these adolescent health behaviors through two specific mechanismsfamily cohesion and problem-focused coping.

Family cohesion is characterized by a feeling of closeness among

family members, enjoyment of family activities, and a style of decision making in which all family members feel involved (Olson et al.,
1985). Family cohesion likely helps girls feel more connected to
family members and willing to turn to them when dealing with
adolescent concerns. Smoking was less likely to occur in families who
ate together and were high in cohesiveness. This finding suggests that
family cohesion may help adolescents to make better choices in
regard to their health behavior, at least in relation to tobacco use.
Studies have consistently found that parental behaviors in regard to
smoking (e.g., parental use), as well as attitudes, can influence adolescent smoking patterns (Chassin et al., 2005; Maurer, Brunson, &
Pleck, 2003). Our data suggest that family meals in childhood, mediated by levels of family cohesion, had a protective impact on
adolescent smoking behavior. Whether this occurs because health
issues are discussed during family meals, or because greater cohesion
is linked in general with more healthful parental/familial behaviors, is
not known. However, our data do suggest that parents should be
encouraged to eat with their young children on a regular basis, as such
interactions may positively influence risky behavior that is not likely
to occur for many years later.
The association between problem-focused coping and reduced
levels of stress and drive-for-thinness and bulimic symptoms suggests positive effects of dealing with problem situations actively
(Koff & Sangani, 1997). By addressing problems head on, girls
may experience less stress than if problems are internalized. Drivefor-thinness and bulimic symptoms may reflect aberrant, internalized ways of dealing with societal pressures to maintain a certain
body shape or weight (Rowe, Pickles, Simonoff, Bulik, & Silberg,
2002). Girls with a tendency to use problem-focused coping may
use more healthy ways of dealing with these societal pressures.
Alternatively, family meals may provide the opportunity for girls
to see positive coping modeled by parents (taking action to solve
problems).
There was no evidence that emotion-focused coping mediated
the relationship between family meals and outcomes measured in

S116

FRANKO, THOMPSON, AFFENITO, BARTON, AND STRIEGEL-MOORE

Year 10. This may be related to the fact that our outcomes were
more problem specific (e.g., smoking) than emotional in nature
(e.g., how to deal with negative affect). Alternatively, discussions
at family meals may focus more on action-oriented problem solving rather than appropriate expressions of emotions in the face of
adversity.
It is interesting that we did not find some of the relationships
between family meals and outcomes that were observed in previous studies with cross-sectional samples, particularly extreme
weight control behaviors (e.g., vomiting and laxative use;
Neumark-Sztainer et al., 2004) and alcohol use (Eisenberg et al.,
2004). This is not surprising given that concurrent associations
between family meals and outcomes are likely to be stronger than
prospective relations; in our study, there was a large gap in time
between measurement of family meals (Years 13) and outcomes
(Year 10). Furthermore, adjusting for earlier levels of the outcomes (Years 5/6) forced us to model differences among girls who
started at the same level of the outcome, which means there is less
variation among girls to explain than there is when such controls
are absent. This study, unlike prior studies, provided stronger
evidence that the frequency of family meals led to positive outcomes and not vice versa.
Limitations of this study are related to the sample composition,
which does not allow us to assess the impact of family meals on
boys or on individuals with other demographic characteristics
(e.g., Hispanic families). The reliance on self-reported data may
also have biased the results in that children might be inclined to
provide desired responses to questions. It is also possible that
children answered positively to the family meals question in situations that might not be ideal (e.g., eating together as a family in
the car while driving to soccer practice). However, our validation
analyses suggested that children did respond to the question in the
spirit in which it was asked, that is, sitting down together and
eating a meal with family. In the future, a more detailed assessment of family meals would provide a fine-grained and complete
picture of the associations. Nevertheless, a single question assessing the frequency of family meals was found to be significantly
related to adolescent health, mediated by family cohesion and
coping skills. Two final limitations related to the number of tests
run on this large sample, which increased the risk of Type I errors
and the imputation of missing data. To reduce the likelihood of
making Type I errors, an alpha level of 0.01 was used for all
associations that were not predicted a priori based on past literature; however, this does not guarantee that Type I errors were not
made. The results should be viewed as hypothesis generating, and
further confirmation should be sought in future studies. A growing
literature on the effects of methods for handling missing data (see
Allison, 2002, for a review) consistently shows that when data are
missing, simply dropping cases with missing data (complete-case
analysis) often leads to biased estimates. Furthermore, estimates
based on techniques such as multiple imputation and hot decking
are less subject to bias, and some studies indicate that hot decking
can be especially advantageous (Tang, Song, Belin, & Unutzer,
2005). In this study, complete-case analysis was compared with
hot decking and the results differed in only one case (i.e., the
association between family cohesion and smoking).
We also note that the clinical significance of some of our
findings is unclear. Due to the large sample size (N 2,379), the
study had excellent power to detect statistical associations between

the variables of interest; however, the effect sizes were consistently small ( [r] 0.14). For example, the estimated coefficient
for the association between problem-focused coping and the stress
scale ( 0.25, SE 0.05) was highly significant from a
statistical perspective ( p .0001), but this translated into a .25
decrease in stress for a 1-point increase in problem-focused coping. Given that stress scores ranged from 0 to 56 (Table 1), the
clinical significance of this decrease is debatable. It is possible that
large increases in problem-focused coping would be needed to
produce clinically significant decreases in stress. Future work is
needed to determine the practical significance of these associations
and whether they predict meaningful differences in health and
well-being.
An important strength of NGHS is the longitudinal nature of the
data. To determine the causal effect of family meals, ideally one
would randomize girls to family meal conditions in an experiment.
Obviously, such an experiment would be difficult and potentially
unethical to implement. The best alternative is to examine prospective relations between family meals and outcomes in a longitudinal sample where natural variation in frequency of family
meals is observed. Although causality cannot be determined with
certainty, at least the present study was able to provide evidence
for a necessary (but not sufficient) condition of causality, namely
that cause precedes effect in time (and effect cannot precede
cause). Because girls were not randomized to family meal conditions, it was important to adjust for possible differences among
groups as well as for the outcome variable at the prior time. As we
were able to conduct these analyses, we can be more confident of
the direction of the relations between variables. However, we
cannot ultimately rule out the possibility of other variables (not
measured) that might be causing the positive outcomes.
In sum, our findings indicate that several important adolescent
health concerns, namely smoking, perceived stress levels, drive for
thinness, and bulimic symptoms, were influenced by the frequency
of family meals in childhood and mediated by family cohesion and
coping styles. These data suggest that families should be encouraged to eat together from an early age, as such activity may have
multiple benefits in later years when adolescent health issues
become paramount.

References
Allison, P. D. (2002). Missing data. Thousand Oaks, CA: Sage.
Borowsky, I. W., Ireland, M., & Resnick, J. D. (2001). Adolescent suicide
attempts: Risks and protectors. Pediatrics, 107, 485 493.
Chassin, L., Presson, C. C., Rose, J., Sherman, S. J., Davis, M. J., &
Gonzalez, J. L. (2005). Parenting style and smoking-specific parenting
practices as predictors of adolescent smoking onset. Journal of Pediatric
Psychology, 30, 333344.
Cohen, S., Kamarck, T., & Mermelstein, R. (1983). A global measure of
perceived stress. Journal of Health and Social Behavior, 24, 385396.
Compton, M. T., Thompson, N. J., & Kaslow, N. J. (2005). Social environment factors associated with suicide attempt among low-income
African Americans: The protective role of family relationships and
social support. Social Psychiatry & Psychiatric Epidemiology, 40, 175
185.
Cook, T. D., & Campbell, D. T. (1979). Quasi-experimentation: Design
and analysis issues for field settings. Boston: Houghton Mifflin.
Cooper, P. J., Whelan, E., & Woolgar, M. (2004). Association between
childhood feeding problems and maternal eating disorder: Role of the
family environment. British Journal of Psychiatry, 184, 210 215.

MEDIATORS OF FAMILY MEALS

Eisenberg, M. E., Olson, R. E., Neumark-Sztainer, D., Story, M., &
Bearinger, L. H. (2004). Correlations between family meals and psychosocial well-being among adolescents. Archives of Pediatrics & Adolescent Medicine, 158, 792796.
Fonseca, H., Ireland, M., & Resnick, M. D. (2002). Familial correlates of
extreme weight control behaviors among adolescents. International
Journal of Eating Disorders, 32, 441 448.
Franko, D. L., Striegel-Moore, R. H., Barton, B. A., Schumann, B. C.,
Garner, D. M., Daniels, S. R., Schreiber, G. B., & Crawford, P. B.
(2004). Measuring eating concerns in black and white adolescent girls.
International Journal of Eating Disorders, 35, 179 189.
Fulkerson, J. A., Neumark-Sztainer, D., & Story, M. (2006). Adolescent
and parent views of family meals. Journal of the American Dietetic
Association, 106, 526 532.
Fulkerson, J. A., Story, M., Mellin, A., Leffert, N., Neumark-Sztainer, D.,
& French, S. A. (2006). Family dinner meal frequency and adolescent
development: Relationships with developmental assets and high-risk
behaviors. Journal of Adolescent Health, 39, 337345.
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.
Garner, D. M., Olmstead, M. P., & Polivy, J. (1983). Development and
validation of a multidimensional eating disorder inventory for anorexia
nervosa and bulimia. International Journal of Eating Disorders, 2,
1534.
Garrison, C. Z., Waller, J. L., Cuffe, S. P., McKeown, R. E., Addy, C. L.,
& Jackson, K. L. (1997). Incidence of major depressive disorder and
dysthymia in young adolescents. Journal of the American Academy of
Child & Adolescent Psychiatry, 36, 458 465.
Hershberger, P. J. (1990). Self complexity and health promotion: Promising but premature. Psychological Reports, 66, 12071216.
Kiesner, J., & Kerr, M. (2004). Families, peers, and contexts as multiple
determinants of adolescent problem behavior. Journal of Adolescence,
27, 493 495.
Kingon, Y. S., & OSullivan, A. L. (2001). The family as a protective asset
in adolescent development. Journal of Holistic Nursing, 19, 102121.
Koff, E., & Sangani, P. (1997). Effects of coping style and negative body
image on eating disturbance. International Journal of Eating Disorders,
22, 5156.
Kraemer, H. C., & Blasey, C. M. (2004). Centring in regression analyses:
A strategy to prevent errors in statistical inference. International Journal
of Methods in Psychiatric Research, 13, 141151.
Kraemer, H. C., Morgan, G. A., Leech, N. L., Gliner, J. A., Vaske, J. J., &
Harmon, R. J. (2003). Measures of clinical significance. Journal of the
Academy of Child and Adolescent Psychiatry, 42, 1524 1529.
Kraemer, H. C., Stice, E., Kazdin, A., Offord, D., & Kupfer, D. (2001).
How do risk factors work together? Mediators, moderators, and independent, overlapping and proxy risk factors. American Journal of Psychiatry, 158, 848 856.
MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., &
Sheets, V. (2002). A comparison of methods to test mediation and other
intervening variable effects. Psychological Methods, 7, 83104.
Mamun, A. A., Lawlor, D. B., OCallaghan, M. J., Williams, G. M., &
Najman, J. M. (2005). Positive maternal attitude to the family eating
together decreases the risk of adolescent overweight. Obesity Research,
13, 14221430.
Maurer, T. W., Brunson, L., & Pleck, J. H. (2003). Adolescent smoking

S117

behavior: The relative influence of parental and peer norms. Adolescent

& Family Health, 3, 130 139.
Nagelkerke, N. J. D. (1991). A note on a general definition of the coefficient of determination. Biometrika, 78, 691 692.
National Heart, Lung, and Blood Institute Growth and Health Study
Research Group. (1992). Obesity and cardiovascular disease risk factors
in black and white girls: The NHLBI Growth and Health Study. American Journal of Public Health, 82, 16131620.
Neter, J., Wasserman, W., & Kutner, M. H. (1990). Applied linear statistical models. Boston, MA: Irwin.
Neumark-Sztainer, D., Hannan, P. J., Story, M., Croll, J., & Perry, C.
(2003). Family meal patterns: Associations with sociodemographic characteristics and improved dietary intake among adolescents. Journal of
the American Dietetic Association, 103, 317322.
Neumark-Sztainer, D., Wall, M., Story, M., & Fulkerson, J. A. (2004). Are
family meal patterns associated with disordered eating behaviors among
adolescents? Journal of Adolescent Health, 35, 350 359.
Olson, D. H., Portner, J., & Lavee, Y. (1985). FACES III. Family Social
Science, University of Minnesota, St. Paul, MN.
Resnick, M. D., Harris, L. J., & Blum, R. W. (1993). The impact of caring
and connectedness on adolescent health and well-being. Journal of
Pediatric Child Health, 29, S39.
Rodriguez Martin, A., Novalbos Ruiz, J. P., Martinez Nieto, J. M., Escobar
Jimenez, L., & Castro de Haro, A. L. (2004). Epidemiological study of
the influence of family and socioeconomic status in disorders of eating
behaviour. European Journal of Clinical Nutrition, 58, 846 852.
Ross, C. E., & Mirowsky, J. (1999). Refining the association between
education and health: The effects of quantity, credential, and selectivity.
Demography, 36, 445 460.
Rowe, R., Pickles, A., Simonoff, E., Bulik, C. M., & Silberg, J. L. (2002).
Bulimic symptoms in the Virginia twin study of adolescent behavioral
development: Correlates, comorbidity, and genetics. Biological Psychiatry, 51, 172182.
Scott, J., Kraemer, H. C., Essex, M. J., & Kupfer, D. J. (2004). Schematic:
A program for classifying and diagramming mediators, moderators, and
independent, proxy and overlapping risk factors. Pittsburgh, PA: Mental
Health Intervention Research Center for the Study of Mood and Anxiety
Disorders, University of Pittsburgh Department of Psychiatry/WPIC.
Tang, L., Song, J., Belin, T. R., & Unutzer, J. (2005). A comparison of
imputation methods in a longitudinal randomized clinical trial. Statistical Medicine, 30, 21112128.
Taveras, E. M., Rifas-Shiman, S. L., Berkey, C. S., Rockett, H. R. H.,
Field, A. E., Frazier, A. L., Colditz, G. A., & Gillman, M. W. (2005).
Family dinner and adolescent overweight. Obesity Research, 13, 900
906.
Tepper, R. L. (1999). Parental regulation and adolescent time-use decisions: Findings from the NLSY97. In R. T. Michael (Ed.), Social
awakening: Adolescent behavior as adulthood approaches. (pp. 77
103). New York NY: Russel Sage Foundation.
Tobin, D. L., Holroyd, K. A., & Reynolds, R. V. (1984). Users manual for
the Coping Strategies Inventory. (pp. 10 11). Ohio University: Department of Psychology.
Tobin, D. L., Holroyd, K. A., Reynolds, R. V., & Wigal, J. K. (1989). The
hierarchical factor structure of the coping strategies inventory. Cognitive
Therapy & Research, 13, 343361.
Wear, R. W., & Pratz, O. (1987). Test-retest reliability for the Eating
Disorder Inventory. International Journal of Eating Disorders, 6, 767
769.