THE COUNSELING

Striegel-Moore, Cachelin
PSYCHOLOGIST
/ ETIOLOGY OF
/ September
EDS 2001
Etiology of Eating Disorders in Women
Ruth H. Striegel-Moore
Wesleyan University
Fary M. Cachelin
California State University, Los Angeles

Eating disorders have been studied extensively over the past several decades, yet
research of their etiology has lagged behind treatment outcome research. This article
reviews the challenges inherent in this research. It illustrates the epidemiologic designs
that have been used to test risk factor hypotheses and describes the major studies
designed to answer the question of what causes eating disorders. It points to significant
gaps in knowledge, chief among them the absence of representative data on prevalence
and correlates of eating disorders, and the lack of data regarding eating disorders in eth-
nic minority populations.

Eating disorders (EDs) pose a considerable threat to the health and well-
being of adolescent girls and adult women. Compared to other major psychi-
atric disorders, EDs have been introduced into the psychiatric nomenclature
relatively recently and, consequently, research is less advanced and major
gaps of knowledge remain. Anorexia nervosa (AN) was described as a new
disease in the late 19th century by the British psychiatrist Gull (1888) and the
French physician Laseque (1873/1964). Both shared the view that AN was a
“nervous disease,” yet their clinical descriptions were quite different. Gull
focused extensively on the physiological correlates of the disorder, asserting
that this disorder involved “simple starvation.” He attributed the disorder to a
“perversion of the will” without paying attention to the causes of this “perver-
sion” or the psychological symptoms that factor prominently into today’s
classification of AN. Laseque classified AN as a “hysteria of the gastric cen-
ter.” He proposed that “emotional distress” played a major causal role and
that family dynamics added further to the risk for or maintenance of the disor-
der. Despite these differences, treatment by both physicians focused on nutri-
tional rehabilitation, essentially reducing a mental disorder to a somatic
problem. The formulation of AN as a medical disease remained dominant
until Hilde Bruch (1973, 1978) introduced her biopsychosocial theory that
emphasized the role of developmental factors and family dynamics.

Ruth H. Striegel-Moore was supported in part by a grant from the National Institute of Mental
Health and the National Institute of Diabetes, Digestive and Kidney Diseases (R01 MH57897).
Correspondence concerning this article should be addressed to Ruth H. Striegel-Moore, Depart-
ment of Psychology, Wesleyan University, 207 High Street, Middletown, CT 06459; e-mail:
[email protected].
THE COUNSELING PSYCHOLOGIST, Vol. 29 No. 5, September 2001 635-661
© 2001 by the Division of Counseling Psychology.
635

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

636 THE COUNSELING PSYCHOLOGIST / September 2001

Subsequently, in a more direct challenge to the medical model, feminist

scholars (e.g., Bordo, 1993; Chernin, 1985; Orbach, 1986) offered a purely
cultural model, describing AN as resulting from women’s subordinate role in
patriarchal society. Although theoretical models of AN proliferated during
the 20th century (for reviews, see Garfinkel & Garner, 1982; Garner &
Garfinkel, 1985), until the 1980s there was little empirical work devoted to
testing etiological models.
When bulimia nervosa (BN) was introduced two decades ago, two com-
peting theoretical models were proposed simultaneously at that time. The
feminist formulation of BN as a culture-bound syndrome attributed risk for
BN to Western societies’ gender role stereotypes, in general, and the over-
valuation of beauty and obsession with thinness in particular (Boskind-
Lodahl, 1976). The medical conceptualization of BN described this syn-
drome as an “ominous variant” of AN (Russell, 1979). The seemingly sudden
appearance of BN reinforced the belief that culture plays a major role in etiol-
ogy. Empirical studies on cultural factors and risk for the development of
EDs, however, have not captured the complex set of cultural variables pro-
posed in theoretical work (e.g., Rodin, Silberstein, & Striegel-Moore, 1985;
Smolak & Striegel-Moore, in press).
Binge Eating Disorder (BED) was included in 1994 in the fourth edition of
the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV)
(American Psychiatric Association, 1994) as a provisional diagnosis in need
of further study. Consistent with the explicit atheoretical stance of recent edi-
tions of the DSM, there is only limited theoretical work regarding BED (e.g.,
Striegel-Moore, 1993). Whether BED will be retained in subsequent editions
of the DSM will depend, in part, on results of etiological research. If it can be
shown that BED is associated with etiological factors that are distinct from
BN, this will strengthen considerably the case for BED as distinct from the
other, already-established ED diagnostic categories.
This article reviews the status of research of the etiology of EDs and pro-
vides suggestions for future research. First, we briefly describe risk factor
terminology. Then, we discuss methodological challenges in risk factor
research. A major focus is on the various designs currently utilized in risk fac-
tor research and the limitations of current findings due to design issues. We
conclude by considering implications of current findings for future research.

DEFINITIONS OF BASIC CONCEPTS

Despite considerable growth in research of EDs over the past two decades,
etiological research still lags behind other productive areas such as treatment
outcome research. There is a voluminous literature that claims to speak to the

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

Striegel-Moore, Cachelin / ETIOLOGY OF EDS 637

question of what factors cause AN or BN, yet closer reading reveals that
many of these studies fail to address the fundamental scientific principle that
to show cause and effect, the causal variable has to precede in time the out-
come variable of interest. As described in detail by Kraemer et al. (1997), eti-
ological research requires that one differentiate between correlates,
concomitants, and risk factors. A correlate is any factor associated with the
outcome. Correlational studies have identified a myriad of variables that are
associated with EDs. These studies have served as the starting point for more
recent research, where methodologies have been used that permit distinction
of concomitants and risk factors. A concomitant is a variable that has been
shown to be associated significantly with the outcome and to not have pre-
ceded the outcome (e.g., low serum potassium level in BN). A risk factor is a
variable that has been shown to precede the outcome (e.g., childhood sexual
abuse). If the risk factor cannot be changed, it is considered a fixed marker
(e.g., sex); if the risk factor can be shown to change spontaneously or to be
changeable, it is considered a variable risk factor (e.g., body image con-
cerns). If it can be shown that manipulation of the risk factor changes the risk
of the outcome, then this variable risk factor is a causal risk factor.
This typology provides a useful framework and has important implica-
tions for risk management (primary and secondary prevention). Fixed mark-
ers are unchanging characteristics and as such are easiest to measure (e.g.,
timing of measurement is not important). Fixed markers typically are identi-
fied long before causal risk factors are. Variable risk factors are more difficult
to measure because they may change. Ideally, such variables should be mea-
sured repeatedly and close in time to when they are likely to occur (to ensure
accurate recall). Data about fixed markers are helpful for developing cost-
effective screening procedures, because fixed markers point to populations
likely to contain cases. Identifying variable risk factors is essential for devel-
oping interventions aimed at reducing the prevalence of EDs.
Previous integrative reviews have summarized the literature that describes
correlates of AN (Bemis, 1978), BN (Striegel-Moore, Silberstein, & Rodin,
1986), and BED (Yanovski, 1993). A recent review summarized the studies
on biological abnormalities in AN and BN (Ferguson & Pigott, 2000). The
authors noted that although there is considerable evidence of alterations in
neurotransmitter and neuroendocrine function in women with EDs, it is not
known whether the abnormalities identified precede the disorder, are conse-
quences of malnutrition and disordered eating, or are permissive factors that
maintain the ED once it has developed. These reviews require no detailed
updates, as subsequent studies have added little new information regarding
correlates. Therefore, the present article primarily will review a second gen-
eration of studies: investigations that provide data regarding risk factors for
EDs. Several methodological challenges inherent to this research, including

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

638 THE COUNSELING PSYCHOLOGIST / September 2001

case definition, low base rates, and multifactorial causation of EDs, are con-
sidered next.

METHODOLOGICAL CHALLENGES
IN RISK FACTOR RESEARCH

Classification of EDs

Since their introduction, criteria for AN and BN have been modified with
every revision of the DSM. These revisions primarily have focused on sever-
ity criteria. As is the case for many other psychiatric disorders, classification
of EDs is complicated by the fact that the core symptoms occur along a sever-
ity continuum, and some of the symptoms (e.g., body image concerns, diet-
ing) are widely prevalent in the female population. The challenge of classifi-
cation is to strike a balance between selecting criteria that are (a) not so
restrictive as to exclude cases that may not evidence all features of the
prototypical expression of the disorder, yet share its essential features; and
(b) not so inclusive as to encompass cases in which the symptom picture does
not appear to be clinically significant. The current DSM-IV diagnostic crite-
ria have been criticized because a majority of individuals who request treat-
ment do not meet criteria for AN or BN (for a review, see Striegel-Moore &
Marcus, 1995). The boundaries that separate EDs from healthy status are
established by severity criteria that have no basis in empirical research. As
described by Kendler (1999), efforts to validate syndromes by use of a “clini-
cal impairment” criterion cannot fully resolve the boundary dilemma
because in practice it proves difficult to establish that dysfunction or disabil-
ity is due to a particular disorder. Moreover, many risk factors for psychiatric
disorders in turn are associated with impairment (e.g., Walker et al., 1999).
The impact of the changing diagnostic criteria on our understanding of risk
factors for EDs has not yet been examined empirically. Even studies with
identical diagnostic criteria have utilized different research instruments to
determine case status. These methodological variations can be expected to
affect diagnosis of BN and BED in particular, because with respect to clinical
features there is considerable overlap between these two disorders (Striegel-
Moore et al., in press). Standardized assessment methods are needed to per-
mit clear comparisons across studies (Regier et al., 1998). Kendler’s group
has illustrated the importance of diagnostic definition for the assessment of
risk: Heritability estimates have varied widely with varying definitions of BN
(for discussion, see Fairburn, Cowen, & Harrison, in press).
The relationships between AN and BN, between AN binge-eating/
purging type and BN, and between BN and BED remain poorly defined. A

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

Striegel-Moore, Cachelin / ETIOLOGY OF EDS 639

considerable subset of individuals with AN eventually develop BN (e.g.,

Bulik, Sullivan, Fear, & Pickering, 1997), yet few etiological studies con-
sider the implications of this phenomenon in terms of risk. One plausible, yet
untested hypothesis is that individuals with pure AN (not comorbid with BN)
or pure BN (no history of AN) represent different groups in terms of etiology
from individuals with the comorbid type. Correlational studies have shown
that individuals with the binge-eating/purging subtype of AN are similar to
individuals with BN in terms of personality traits and psychiatric
comorbidity (e.g., Shisslak, Pazda, & Crago, 1990). Similarly, studies sug-
gest considerable similarity between individuals with nonpurging BN and
those with BED (Striegel-Moore et al., in press). To date, few studies have
been designed to compare these groups in terms of etiology, yet such research
is central (along with research of clinical course and treatment response) to
clarifying the validity and clinical utility of these different diagnostic
categories.

Low Base Rates

The fact that EDs are relatively rare poses a major challenge to risk factor
research. For example, the Epidemiological Catchment Area Study (ECA)
found only 11 cases in a nationally representative sample of more than
20,000 adult women and men (Robins et al., 1984). There is no database in
the United States that would permit an estimate of the prevalence of BN in a
nationally representative sample. (The ECA was initiated before BN was
introduced into the DSM; therefore, screening questions did not capture
bulimic symptoms.) A Canadian study, recruiting a representative sample of
3,831 male and 4,285 female residents of Ontario (aged 15 to 65), identified
only 62 cases of BN (more than 90% of them female) (Garfinkel et al., 1995,
1996). In the Oregon Adolescent Depression Project, to date the largest epi-
demiological study of adolescents in the United States that included diagnos-
tic questions regarding EDs, none of the 2,564 boys met lifetime criteria for
AN and only 1 was diagnosed with a history of BN (DSM-III). Among the
2,544 girls, 12 and 17 met lifetime criteria for AN and BN, respectively
(Lewinsohn, Hops, Roberts, Seeley, & Andrews, 1993). Clearly, population-
based efforts at locating an adequate number of cases require screening very
large samples. The effort and cost in obtaining a representative sample of
people with EDs are substantial.
Because of the low rates of AN and BN among males (only about 10% of
cases of AN or BN are male), data about risk for developing an ED in males
are especially limited and will not be the focus of this article. The gender
imbalance regarding prevalence is less pronounced in BED, yet most studies
of BED also have focused on female populations. There is an interesting par-

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

640 THE COUNSELING PSYCHOLOGIST / September 2001

adox regarding gender and EDs: Despite the clear emphasis on female popu-
lations in research, the question of why women (Striegel-Moore et al., 1986)
are so much more at risk than men has become increasingly less prominent in
theoretical and empirical work. Current risk factor research tends to be pre-
sented without explicit reference to biological sex or gender roles as impor-
tant explanatory variables. Theoretical models and empirical studies of EDs
need to encompass a consideration of risk related to female biological pro-
cesses and the role of gender in the sociocultural context.
Because EDs are relatively uncommon, many studies have relied solely on
patient samples. Sampling biases in use of patients have been shown in epide-
miological studies (e.g., Kessler, Olfson, & Berglund, 1998). To date, only
two studies have examined bias in patient samples (Fairburn, Welch, Nor-
man, O’Connor, & Doll, 1996; Wilfley, Pike, Dohm, Striegel-Moore, &
Fairburn, in press). A study in Oxford, England, found that parental obesity
and parental drug abuse were significantly less common among clinic
patients than among community cases with BN (Fairburn et al., 1996), sug-
gesting sampling bias may influence results regarding the role of certain risk
factors.
The overreliance on patient samples has contributed to the perception that
EDs are limited to European American populations. A recent study found
that a significantly greater number of African American women responded to
a community-based risk factor study than to a treatment study for BED, even
though these studies were conducted concurrently in the same geographic
area and used similar advertisement strategies (Wilfley et al., in press). This
finding suggests that relying solely on patient samples may result in the inad-
vertent exclusion of ethnic minority groups.
EDs are widely seen as disorders of European American females, and girls
or women from ethnic minority groups are believed to be protected from
developing AN or BN (Striegel-Moore & Smolak, 1996). There is emerging
evidence, however, that ethnic minority groups experience ED symptoms
(for reviews, see Crago, Shisslak, & Estes, 1996; Striegel-Moore & Smolak,
2000), and risk factors unique to ethnic minority group status (e.g., accultura-
tion, immigration-related stresses, discrimination) need to be considered in
culturally sensitive models of risk. To date, scientific evidence regarding the
prevalence of EDs (and risk factors for developing AN or BN) is too limited
to permit conclusive statements about risk for EDs among ethnic minority
groups. In light of the low rates of EDs, epidemiological studies will need to
oversample individuals from ethnic minority populations to ensure adequate
statistical power for testing hypotheses regarding ethnicity as a risk factor for
developing an ED.

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

Striegel-Moore, Cachelin / ETIOLOGY OF EDS 641

Risk Is Multifactorial

Numerous theoretical models have been proposed to explain the etiology

of EDs, and these models suggest the following risk domains: the socio-
cultural context (e.g., thin beauty ideal, gender roles), the familial and inter-
personal context (e.g., family dynamics, adequacy of parenting, peer influ-
ences), personal vulnerability factors (including genetic factors, personality
traits, and behaviors) and traumatic life events (e.g., physical and sexual
abuse). Even though models differ in the emphasis placed on given risk
domains, there is agreement that the etiology of EDs is multifactorial; indeed,
many of the models overlap considerably in terms of the particular variables
that are hypothesized to be involved (see Striegel-Moore & Cachelin, 1999).
Measurement of exposure to risk, therefore, requires collecting data pertain-
ing to multiple risk domains and, within domains, multiple risk factors. Mul-
tiple risk domains are investigated in a small number of recent studies, chief
among them the work by Fairburn and his group in Oxford (Fairburn, Cooper,
Doll, & Welch, 1999; Fairburn et al., 1998; Fairburn, Welch, Doll, Davies, &
O’Connor, 1997). Although exceptionally broad in the scope of risk variables
under investigation, this research is relatively limited in terms of biological
risk factors. Conversely, research focusing particularly on biological risk fac-
tors (for review, see Lilenfeld & Kaye, 1998) tends to be limited in its assess-
ment of behavioral or cultural risk factors. Hence, to date, a pressing gap con-
cerns the integration of biological factors with behavioral and cultural
factors.
Ideally, exposure to risk should be measured prospectively; and, because
many of the hypothesized risk factors occur in childhood, whereas onset of
the disorders occurs during adolescence, prospective studies need to follow
participants over long periods of time. Prospective studies published to date,
however, either have encompassed relatively short follow-up periods (e.g.,
Leon, Fulkerson, Perry, & Early-Zald, 1995), been carried out with relatively
small samples and therefore have yielded only a handful of ED cases (e.g.,
Patton, Johnson-Sabine, Wood, Mann, & Wakeling, 1990), or, because of the
low number of cases, focused on disordered eating status or partial syndrome
cases rather than full syndromes (Killen et al., 1994, 1996; Stice, Cameron,
Killen, Hayward, & Taylor, 1999).
Research has yet to investigate the relative contribution of the various risk
factors to the development of EDs and the particular ways in which risk fac-
tors combine or interact to promote disordered eating. Recent studies have
found a “dose-response effect” when examining the cumulative effect of risk
factors. For example, both Taylor and colleagues (1998), and Fairburn and
colleagues (Fairburn, Cooper, Doll, & Welch, 1999; Fairburn et al., 1997,
1998), found that the odds ratios for developing disordered eating or EDs

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

642 THE COUNSELING PSYCHOLOGIST / September 2001

increased in linear fashion from individuals exposed to the fewest risk factors
to those exposed to the most factors.
Investigators have implemented a variety of research designs to test risk
factor models of EDs: cross-sectional population surveys, family studies,
twin studies, case-control studies, and prospective cohort studies. Not all
designs are equally well suited for examining a particular risk domain, and
multiple designs are needed to accommodate the challenges described above
(Zahner, Hsieh, & Fleming, 1995). The next section describes each of these
designs and reviews exemplary studies representing the particular method-
ological approaches. Organizing the literature in terms of the design features
of the risk factor studies allows an emphasis on the specific contributions of
the studies to our understanding of risk. To aid a coherent integration of
research findings within a multifactorial framework, Table 1 summarizes the
major findings by risk domain and study design.

VARIOUS RESEARCH DESIGNS

USED TO IDENTIFY RISK FACTORS

Cross-Sectional Population Surveys

In cross-sectional survey studies, a probability sample of participants is
selected to represent the population to which results are to be generalized.
Survey studies are well suited for gathering information on all cases of disor-
der occurring in a defined population. Because data about disease status and
exposure to risk factors are obtained at the same time, risk factor hypotheses
can be tested only for risk factors for which it can be assumed they preceded
the disease (e.g., sex, ethnicity). For disorders for which treatment seeking is
low or disproportionately less common among certain subgroups (e.g., eth-
nic minorities), survey studies can identify important risk factors that might
be missed in other designs. Because cross-sectional studies typically sample
large groups of participants, they are costly. One strategy to contain cost is to
limit the survey questions. Consequently, most surveys do not measure all
risk domains. The cross-sectional survey design is not well suited for risk fac-
tors for which the temporal sequencing between exposure and disease onset
requires measurement. Nevertheless, in practice, many survey studies gather
data about exposure to risk by retrospective reporting. This information then
provides a valuable starting point for generating hypotheses to be tested in
studies utilizing designs that do not rely on retrospective assessment of risk
(e.g., prospective cohort studies).
During the 1980s, several survey studies were conducted to determine
prevalence and correlates of AN and BN. To date, there is no population sur-

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

Striegel-Moore, Cachelin / ETIOLOGY OF EDS 643

TABLE 1: Risk Factor Variables (by domain) and Corresponding Study Design(s)

Risk Domain Type of Eating Disorder Type of Study Design

Demographic characteristics
Sex AN, BN Survey, family study, twin study
Age AN, BN Survey
Birth cohort BN Survey, prospective cohort
Parental socioeconomic status AN, BN Survey, twin study
Sociocultural context
Acculturation m.s. Survey
Familial interpersonal and context
Frequent house moves n.s. Case-control
Inadequate parenting AN, BN, BED Case-control
Parental psychopathology BN Family studies, case-control
Abuse (sexual, physical) AN, BN, BED Case-control, survey
Conflicted relationships with parents DE Prospective cohort
Parental high expectations BN Case-control
Parental concern with weight/eating AN, BN, BED Case-control
Familial eating disorders AN, BN, BED Family studies, twin studies,
case-control
Parental obesity BN Case-control
No close friends AN, BN Case-control
Bullied by other children BN, BED Case-control
Constitutional vulnerability
Premature birth AN Twin studies
Severe health problems AN, BN Case-control
Childhood obesity/body fat AN, BN, BED Case-control, prospective study
Early onset of menarche BN, DE Case-control, prospective study
Pregnancy (prior to onset) BN, BED Case-control
Perfectionism AN, BN, BED Case-control, prospective study
Personal vulnerability
Psychopathology AN, BN, BED Case-control, prospective study
Negative affectivity BN, BED Case-control, prospective study
Social self disturbance BN Case-control, prospective study
Low self-esteem AN, BN, BED Case-control, prospective study
Internalized thin ideal DE Prospective study
Body image concerns DE Case-control, prospective study
Dieting DE Prospective study

NOTE: AN = anorexia nervosa, BN = bulimia nervosa, BED = binge eating disorder, DE = disor-
dered eating, m.s. = mixed support, n.s. = not significant.

vey of BED in which diagnosis was based on interview assessment, an essen-

tial methodological feature to ensure reliability of diagnoses (Fairburn &
Beglin, 1990). This gap in the literature is significant, given that preliminary
data suggest that BED may be more common than AN or BN (Striegel-
Moore, 2000). With a few noteworthy exceptions (e.g., Lewinsohn et al.,

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

644 THE COUNSELING PSYCHOLOGIST / September 2001

1993; Whitaker, 1992), survey studies are based on adult samples (van
Hoeken, Lucas, & Hoek, 1998). To date, population surveys contain too few
individuals from ethnic minority populations to be useful for answering the
question of whether ethnicity is a risk factor for developing an ED or whether
there are certain risk factors that are unique to some ethnic groups.
Cross-sectional population surveys have reported three consistent find-
ings. One, EDs are relatively uncommon. Two, AN and BN are significantly
more likely to occur in females than in males. Hence, female sex is accepted
as a fixed marker for AN and BN. Three, AN or BN cases tend to be younger
than individuals who do not have an ED, a result that reflects two separate risk
factors: age and birth cohort. Onset of AN and BN has been shown to occur
during adolescence or young adulthood (e.g., Newman et al., 1996), and
younger cohorts have been found to have higher rates of EDs than older
cohorts (e.g., Bushnell, Wells, Hornblow, Oakley-Browne, & Joyce, 1990).
One recent population survey (Garfinkel et al., 1995, 1996) illustrates the
contributions of this design to understanding risk for EDs. By design, this
Canadian study did not select participants on the basis of their disease status.
Hence, an important, unique contribution of this study was that it permitted
comparisons between full-syndrome (FS) and partial-syndrome (PS) cases
regarding exposure to risk factors. Women with FS-AN did not differ signifi-
cantly from women with PS-AN (defined as meeting all criteria except for
amenorrhea) on rates of parental psychopathology or childhood sexual
abuse. The similar risk factor profiles of FS-AN and PS-AN cases supports
those who have called for the elimination of amenorrhea as a defining symp-
tom of AN in women (Cachelin & Maher, 1998; Garfinkel et al., 1996;
Striegel-Moore & Marcus, 1995). Comparisons of women with PS-BN
(defined as meeting all but the binge frequency criterion) and FS-BN also
revealed no significant differences in exposure to the risk factors of parental
psychopathology and childhood sexual abuse between the two groups. Based
on these results, Garfinkel and colleagues (1995) suggested that the fre-
quency criterion of two binge episodes per week was not supported empiri-
cally and that PS-BN and FS-BN reflect different levels of severity of a com-
mon spectrum disorder.
Acculturation becomes a variable of interest and importance when exam-
ining disordered eating in women of various ethnicities. Exposure to and
adoption of Western values concerning attractiveness and thinness have been
proposed as primary risk factors for the development of EDs (e.g., Striegel-
Moore et al., 1986); acculturation level is one way to measure exposure to and
adoption of U.S. social values. To date, research is preliminary (e.g., no study
has used a representative sample) and results are inconsistent. Some
researchers have reported a significant positive relationship between higher
acculturation level and increased disordered eating (e.g., Cachelin, Veisel,

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

Striegel-Moore, Cachelin / ETIOLOGY OF EDS 645

Striegel-Moore, & Barzegarnazari, in press; Davis & Katzman, 1999), yet

other studies have not found such a relationship (e.g., Furukawa, 1994; Joiner &
Kashubeck, 1996). The inconsistent results probably are due to the different
populations studied as well as to the various definitions and measures of
acculturation employed. Research is needed to further investigate the role
acculturation plays in disordered eating among minority women.
In conclusion, population surveys are of undisputed value for document-
ing the extent to which a specific population contains individuals with the
disorder and for testing risk factors that safely can be assumed to precede
onset of the disorder. The most pressing gap in the literature to date for which
this design is suitable concerns the question of prevalence and correlates of
EDs in the United States. Furthermore, data are needed to determine whether
ethnicity is a risk factor for the development of EDs.

Family Studies
Family studies represent the first step in a systematic progression of
genetic epidemiologic research, permitting an answer to the question of
whether a disorder is familial (i.e., runs in families). Evidence of familial
aggregation then leads to a series of subsequent questions (requiring different
designs), such as the extent to which the transmission can be attributed to
genetic versus environmental factors and what mechanisms underlie the
familial aggregation. The family study design involves identifying a case
sample (probands), a control sample of individuals who do not have the target
disorder, and a corresponding sample of relatives (most commonly first-
degree relatives) of these probands and their controls. By interviewing rela-
tives directly to ascertain psychiatric diagnoses, the family study generates
more accurate estimates of familial aggregation than the family history
method (which involves interviewing the probands about disorders in their
relatives). Moreover, controls must be recruited from the same population as
the probands. This requirement can be difficult to achieve, given that many
family studies are carried out with probands who are patients at university-
based treatment centers that attract patients from well beyond the immediate
community surrounding the treatment center. Last, interviewers conducting
the assessment of the relatives need to be blind regarding the diagnostic status
of the proband.
A review of family studies illustrates that with one exception (involving a
very small sample), each study found more first-degree relatives with an ED
among the families of probands with AN or BN than among the relatives of
control probands (Lilenfeld & Kaye, 1998). An important finding of these
family studies concerns the coaggregation of AN and BN among the relatives
of AN and BN probands. Specifically, these studies have shown that among

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

646 THE COUNSELING PSYCHOLOGIST / September 2001

AN probands a significantly greater number of relatives have BN than

observed in the families of control probands; likewise, more relatives with
AN are found among BN probands than among the family members of con-
trol probands. These findings have been confirmed in a well-controlled new
family study involving the largest case series of patients with EDs used in any
published etiological study to date. Specifically, Strober, Freeman, Lampert,
Diamond, and Kaye (in press) determined prevalence of EDs in the 1,561
first-degree relatives (aged 12 and older) of 152 AN probands (restricting
type), 171 probands with BN, and 181 control probands. AN and BN were
defined as “pure types,” screening out any anorexic individuals with a history
of BN and any bulimic individuals with a history of AN. Control probands
were identified using an acquaintance method, a cost-efficient means of
recruiting demographically matched control probands. Control probands
were screened for history of a psychiatric disorder and were excluded if they
met lifetime criteria. ED assessment in relatives permitted both full and par-
tial syndrome diagnoses. Three main findings were observed. One, EDs were
significantly more common in relatives of probands with AN or BN than in
relatives of control probands. Risk for AN in relatives of probands with AN
was 11.4 times as high as the risk in relatives of control probands; risk for BN
was 3.7 times as high among relatives of BN probands than among control
probands. All relatives with EDs were female. Two, coaggregation was con-
siderable. Specifically, the age-corrected risk for BN was 3.5 times as high in
the relatives of anorexic probands as in the relatives of control probands, and
the risk for AN among relatives of BN probands was 12.1 times as high.
Three, partial syndromes also showed a pattern of aggregation, although less
pronounced than observed for full syndromes. These findings support the
notion that AN and BN have common etiological factors and suggest etiolog-
ical continuity between full and partial syndromes. Strober et al. highlighted
one further, unexpected, finding: None of the relatives with an ED met crite-
ria for both AN and BN. According to the investigators, one possible inter-
pretation of this result is that pure ED probands may represent an etiologi-
cally distinct type of AN or BN compared to individuals who “cross over”
from one to the other ED.

Twin Studies
Twin studies offer the opportunity of exploring the genetic and environ-
mental contributions to the etiology of EDs. Because monozygotic (MZ)
twins are genetically identical, discordances regarding a trait or disorder
observed in MZ twins can be attributed to environmental factors. Dizygotic
(DZ) twins, in contrast, share no more genetic similarity than nontwin sib-
lings. The goal of the classic twin study is to compare similarities and differ-

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

Striegel-Moore, Cachelin / ETIOLOGY OF EDS 647

ences between MZ and DZ twins to determine genetic and environmental

causal factors for EDs. In a recent review, Bulik and her colleagues (Bulik,
Sullivan, Wade, & Kendler, 2000) described in detail the assumptions that
have to be met for genetic studies to generate valid data regarding the contri-
bution of genetic factors to the etiology of disorders. These assumptions
include that zygocity is measured accurately, twins and singletons (non-
twins) are comparable regarding the trait under study, and the exposure of
MZ and DZ twins to environmental influences of relevance to the risk for the
disorder under study is equal for MZ and DZ twins (i.e., equal environment
assumption). Reliable methods for determining zygocity make it possible for
investigators to meet the first assumption. In contrast, the assumption of com-
parability of twins to singletons on relevant variables has not yet been consid-
ered fully either conceptually or empirically. For example, pregnancy and
delivery of twins differ from those of singletons (e.g., twins tend to have
shorter gestation periods, more complications during delivery, and lower
birth weights). Whether these differences are of relevance the risk of develop-
ing an ED remains to be examined more fully. A recent population-based
study in Sweden took advantage of the availability of two databases: a
national birth register and a national inpatient register. Using 781 girls who
had been hospitalized for AN and 3,905 controls matched to cases on age and
hospital of delivery, this study found that girls born very preterm (gestational
age of 32 weeks or less) had a threefold increase in risk for AN (Cnattingius,
Hultman, Dahl, & Sparen, 1999). This study suggests that parameters related
to pregnancy and delivery may contribute to risk for AN. Finally, the equal
environment assumption as it relates to risk for developing an ED has been
tested in only a limited way to date and with conflicting results (Bulik,
Sullivan, & Kendler, 1998; Hettema, Neale, & Kendler, 1995; Klump, Holly,
Iaccoma, McGue, & Wilson, 2000; Sullivan, Bulik, & Kendler, 1998).
Initial twin studies relied on twins recruited among patient samples and
tended to include small samples of twin pairs (for review, see Lilenfeld &
Kaye, 1998). Concordance rates and heritability estimates varied widely, and
this early literature is difficult to interpret because of the use of varying diag-
nostic criteria, overreliance on patient samples, and the inadequate power for
testing hypotheses about genetic liability. More recent twin studies have uti-
lized twin registries to gain access to large, community-based samples of
twins. Walters and Kendler (1995) explored the genetic and environmental
contributions to risk for AN in a sample of 590 MZ twins and 440 DZ twins
from the Virginia Twin Registry (VTR). The concordance rates for AN
(broadly defined) in MZ and DZ twins were 10% and 22%, respectively. As
the authors pointed out, the higher rate of concordance for AN among DZ
twins was unexpected and not supportive of a genetic explanation of the dis-
order. Walters and Kendler emphasized that their study did not have enough

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

648 THE COUNSELING PSYCHOLOGIST / September 2001

power to permit partitioning of genetic and environmental factors or estimat-

ing accurately the heritability of AN.
Data from the VTR regarding the heritability of BN have been analyzed
and reported multiple times, with varying results (for review, see Fairburn
et al., in press). In the initial report (Kendler et al., 1991), concordance rates
for BN observed in MZ and DZ twins were 22.9% and 8.7%, respectively.
The authors concluded that although model fitting found that about 50% of
the liability for BN was due to genetic factors, an alternative model in which
twin resemblance was assumed to be due to familial-environmental factors
was almost as good a fit. Kendler and colleagues further indicated that assum-
ing that genetic factors were significant, the study had insufficient power to
detect an additional modest family-environmental effect.
Reanalyses of the VTR data have taken advantage of improved diagnostic
assessment (thus reducing error variance and improving power): Reassess-
ment of BN cases found that the test-retest reliability of the diagnosis of BN
was very low. Accounting for this source of error, more recent reports (Bulik
et al., 1998; Kendler et al., 1991) have concluded that BN is highly heritable.
The methodological aspects of this work (e.g., the particular case definitions,
the use of the tetrachoric correlation coefficient as a statistical measure of
twin concordance) have been debated (Fairburn et al., in press; Kraemer,
1997), and more research is needed to resolve these methodological issues.
In conclusion, family studies and genetic studies have found consistently
that EDs aggregate in families. The contribution of genetic factors relative to
environmental factors in AN has been difficult to establish, given the low
number of twins with AN. Work by Kendler’s group (Bulik et al., 1998;
Kendler et al., 1991; Walters & Kendler, 1995) suggests that genetic factors
play a considerable role in the etiology of BN, but important methodological
questions remain to be resolved. Genetic studies may shed light on the contri-
bution of shared and nonshared environmental factors (after all, heritability
estimates are less than 100%), yet the potential of genetic studies to inform us
about environmental factors has not been fully realized. For example, mea-
surement of environmental factors has been rather limited in the twin studies
published to date. Moreover, research of nonaffected MZ twins should focus
on the question of why they were protected from developing the disorder.

Case-Control Studies
The case-control design involves selecting participants who do (cases)
and who do not (controls) have the disorder of which the etiology is to be
studied. Controls need to be sampled from the same population from which
the cases arise and need to be matched on variables known to affect exposure
history (e.g., gender, age, socioeconomic status). The two groups are com-

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

Striegel-Moore, Cachelin / ETIOLOGY OF EDS 649

pared with respect to existing or past characteristics judged to be of possible

relevance to etiology. This design is recommended for the study of rare dis-
eases. Several limitations need to be considered. The selection of community
cases is preferable to inclusion of patient cases; however, in the absence of
comprehensive population registries of individuals with psychiatric disor-
ders, in practice the process of identifying community cases is comparable to
a population survey in cost and effort. Timing of exposure must be measured
and given the retrospective nature of this assessment; the reliability of these
data may be compromised due to memory loss and cognitive distortions. Par-
ticipants’ current emotional state (e.g., depression) may influence memory of
past events. Hence, among cases, accuracy of recall may be particularly prob-
lematic. Moreover, cases may report more readily exposure to experiences
that are consistent with their current symptoms. For example, the preoccupa-
tion with weight and shape that characterizes individuals with BN may
make it more likely that they report having been teased about their weight
(Vitousek & Hollon, 1990).
Fairburn and colleagues (1997, 1998, 1999) conducted a series of risk fac-
tor studies, recruiting from the registries of general practitioners in
Oxfordshire (England) cases of AN, BN, and BED, along with two types of
controls (healthy, psychiatric). A critically important design feature was that
the time frame for assessment of exposure to risk factors in the control groups
was “yoked” to the cases. Specifically, exposure history in cases was limited
to the time preceding onset of clinically significant symptoms of the ED.
Similarly, exposure in the controls was limited to the time before the onset
age of the case for which they served as the control. Risk factor assessment
encompassed 47 factors from the following domains: personal vulnerability
(childhood obesity, personality traits, childhood psychopathology), familial
and interpersonal factors (family demographic characteristics, parental
psychopathology, family interaction patterns, quality of parenting, family
eating patterns, peer relationships, social pressures about weight and shape,
having a close friend), and adverse life events (childhood physical and sexual
abuse).
The risk factor study for BN included 102 women with BN and (individu-
ally matched to the cases on age and education) 204 healthy controls and 102
psychiatric controls (Fairburn et al., 1997). Multivariate analyses identified
childhood and parental obesity, premorbid psychiatric disorder, poor
parenting, parental psychiatric disorders, and physical and sexual abuse as
significant distinguishing factors when comparing the bulimic women and
their matched healthy controls. There appeared to be a cumulative risk effect:
The greater the number of risk factors reported, the greater was the risk for the
development of BN.

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

650 THE COUNSELING PSYCHOLOGIST / September 2001

In comparisons of cases with psychiatric controls, few risk factors were

found that were uniquely associated with BN (specific risk factors) rather
than with experiencing a psychiatric disorder (common risk factors). Spe-
cifically, only obesity risk and high parental expectations remained signifi-
cant factors differentiating women with BN from women with other psychi-
atric disorders. Hence, this study suggests that personal vulnerability
(reflected in premorbid low self-esteem, depression, and anxiety) combined
with inadequate parenting, parental psychopathology, and physical and sex-
ual abuse set the stage for the development of a psychiatric disorder. Risk spe-
cifically for BN arises when, added to this constellation of risk factors, a per-
son experiences childhood obesity and familial pressure to achieve
excellence. An unexpected result was that parental ED was not found to be a
specific predictor of risk for BN. This negative result may be due to the fact
that the case-control design is not efficient for studying rare exposures.
Moreover, in the Oxford study, parental disorder was ascertained by inter-
viewing the BN cases about family psychiatric disorders. This approach has
been shown to underestimate the rates of disorders in the family (Zahner
et al., 1995), thus further compounding the low base rate of this risk factor.
The risk factor study of AN included 67 women with AN (due to the low
rate of AN in the community, a majority of these cases were recruited from
clinical settings). Data from the 204 healthy controls and 102 psychiatric
controls who participated in the risk factor study for BN were used. Controls,
therefore, were not matched to cases on age or social class or on onset age;
rather, these variables were used as covariates in risk factor analyses.
Compared to healthy controls, women with AN were significantly more
likely to report perfectionism, low self-esteem, severe personal health prob-
lems, a history of deliberate self-harm, major depression, and premorbid
drug abuse. Significant familial risk factors included inadequate parenting,
frequent house moves, high parental expectations, and a parental history of
AN or BN. Women with AN were more likely than healthy controls to report
that they had no close friends. Physical and sexual abuse were more com-
monly reported among women with AN than among healthy controls. Risk
was additive: Odds ratios increased, in a linear fashion, from women exposed
to the smallest number of factors to those exposed to the largest.
Impulsive behaviors such as drug abuse or deliberate self-harm have been
reported to be rare in women with AN, especially in those with the restricting
type of the disorder (e.g., Strober, 1995). Nevertheless, in this case-control
study, premorbid drug abuse and self-harm were found to be (nonspecific)
risk factors for AN. Fairburn and colleagues (1999) did not provide a detailed
description of the AN sample. A majority of women with AN in this study
were recruited from a psychiatric patient registry. Perhaps women with AN

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

Striegel-Moore, Cachelin / ETIOLOGY OF EDS 651

who receive psychiatric treatment are particularly likely to have a history of

impulsive behaviors and, therefore, inadvertently, the Oxford study may have
oversampled women with AN with histories related to impulsive behavior.
Comparisons between women with AN and psychiatric controls revealed
only two distinguishing factors: The women with AN were more likely to
report negative self-evaluation and perfectionism prior to onset. Parental ED
was not found to be a specific risk factor for AN. This result was unexpected
and may point to the weakness of this design in identifying certain risk fac-
tors. Furthermore, given the emphasis placed on cultural factors (e.g., thin-
ness ideal) in some theoretical models of AN, it is perhaps surprising that
familial criticism about weight or family dieting did not prove to be specific
risk factors for AN. However, these results resonate with theoretical models
of AN that emphasize self deficits as core features of the disorder (Bruch,
1978; Strober, 1995).
Comparisons of women with AN and women with BN found only three
differences: Parental psychopathology, parental obesity, and family dieting
or weight concerns were reported significantly more often among women
with BN. In the absence of more detailed description of the clinical character-
istics of the AN and BN samples, it is not clear whether the considerable simi-
larity in risk profiles is due to the fact that there was considerable diagnostic
overlap. For example, it is not reported how many of the women with BN had
a history of AN and how many of the women with AN exhibited binge eating
and/or purging. The two groups also differed in age, socioeconomic status,
and age of onset of the ED. The impact of these differences on risk analyses is
difficult to determine.
In the Oxford study, BED was defined using a more inclusive frequency
criterion (average number of 1 day per week with at least one binge episode)
than has been recommended in the DSM (average number of 2 days per
week). Fifty-two women with BED were compared to 104 matched healthy
controls and 102 psychiatric controls in regard to exposure to the risk factors
measured in the BN study. The 102 psychiatric controls represent the same
women who had been included in the risk factor study for BN and AN. For
psychiatric controls, age, education, and onset age were not matched to the
BED cases. In a multivariate model determining risk for BED relative to
healthy controls, parental psychopathology, along with inadequate parenting
and social pressure about weight and shape, were significant predictors of
risk for BED. Hence, results suggest a combination of a difficult family envi-
ronment and social pressure about weight and shape as increasing risk for
BED. None of the personal vulnerability factors were found to be significant
in predicting risk for BED. Multivariate comparisons of women with BED
and psychiatric controls found that only childhood obesity differentiated the

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

652 THE COUNSELING PSYCHOLOGIST / September 2001

two groups. These results need to be interpreted cautiously, given that the two
groups were different in regard to age, education, and the time period used for
establishing exposure to risk.
There were no significant differences between the women with BED and
women with BN on any of the various risk factors under investigation; how-
ever, women with BN were exposed to a greater number of risk factors than
women with BED. If this finding is replicated in other studies, it supports the
view that BED is not distinct in etiology from BN.
In summary, case-control studies have provided support for a number of
risk domains contributing to the development of EDs. This design involves
retrospective data about risk exposure; hence, results need to be confirmed in
prospective studies. Research needs to examine how multiple risk factors
interact to cause EDs and whether these risk factors hold for ethnic minority
women.

Prospective Cohort Studies

Prospective analyses are helpful in identifying the risk variables in the
development of particular types of psychopathology before there is a con-
founding with the clinical effects of the disorder itself, and they can shed light
on mechanisms that may influence development of EDs and identify vari-
ables that could serve as potential targets for prevention efforts. In prospec-
tive studies, too few ED cases have been identified, and, consequently, these
studies typically use disordered eating (definitions vary among studies) or
partial syndrome EDs as the dependent variable. Several of these studies are
highlighted below.
Graber, Brooks-Gunn, Paikoff, and Warren (1994) assessed 116 adoles-
cent girls (mean age 14.3) three times over an 8-year period from young ado-
lescence to young adulthood. The sample consisted of a nonclinical group of
girls who, at study entry, exhibited a range of eating behaviors from healthy
to problematic. Based on eating attitudes, dieting behaviors, and bulimic
behaviors, the sample was divided into four risk trajectories: chronic eating
problems (high risk), low risk, early-transient risk, and late-transient risk.
Variables of interest were physical development (height, weight, percentage
of body fat, age at menarche), body image (satisfaction with body parts and
appearance), self-image and psychological well-being (depressive affect,
feelings of ineffectiveness, perfectionism, aggressive affect, delinquency,
hyperactivity), and family relationships (family cohesiveness, conflict, and
warmth). Results suggest that entry onto a chronic-risk trajectory for eating
problems was associated with several personal vulnerability factors such as
overweight and early maturational timing, negative body image, and
psychopathology (particularly depressive affect); eating problems had only a

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

Striegel-Moore, Cachelin / ETIOLOGY OF EDS 653

minimal association with family relations (only in young adolescence). Lim-

itations of this study were the select nature of the sample (mainly upper-
middle-class White girls), the small sample size, and the reliance on self-
report questionnaires. This research was important in demonstrating that the
development of eating problems appears to follow different trajectories dur-
ing the adolescent decade and that use of longitudinal analyses can clarify
these paths in ways that could not be achieved with cross-sectional examina-
tions (Graber et al.).
Killen et al. (1994) conducted a prospective analysis of onset of ED symp-
toms in a community sample of 939 adolescent girls over a 3-year interval.
Mean age at baseline was 12. Variables examined were weight concerns (fear
of weight gain, worry over weight and shape, importance of weight, diet his-
tory, and perceived fatness), dietary restraint, pubertal development, Body
Mass Index (BMI), behavior problems (aggressiveness and unpopularity),
and psychological and behavioral characteristics of EDs (including perfec-
tionism, maturity fears, interoceptive awareness). Of these risk variables,
only the measure of weight concerns was significantly associated with onset
of partial BN. This study provided the first prospective demonstration of a
linkage between weight and body shape concerns and later onset of ED
symptoms (Killen et al., 1994).
Killen and colleagues (1996) again examined factors prospectively asso-
ciated with onset of partial syndrome EDs (observed among 36 girls; 4% of
the sample) over a 4-year interval in a different community sample of 877
high-school adolescent girls (mean age 14.9). The sample in this study was
more ethnically diverse than in prior research, consisting of 46% White, 14%
Hispanic, 23% Asian, 6% Pacific Islander, 3% Black, 2% Native American,
and 6% other. (The authors did not report analyses by ethnic group.) The fac-
tors assessed were similar to those examined in the earlier study (Killen et al.,
1994), with the addition of measures of self-reported frequency of alcohol
consumption and temperamental traits including emotionality (distress, fear,
anger), activity, and sociability. Once again, in multivariate analyses only the
measure of weight concerns was significantly associated with onset of partial
syndrome BN. The researchers concluded that weight concerns are an impor-
tant factor in the development of disordered eating during adolescence
(Killen et al., 1996). It remains to be determined, however, whether factors
that contribute to development of partial-syndrome EDs are the same as risk
factors for the development of full-syndrome EDs. Also, it is unclear from
this research whether risk factors for BN are the same as those for AN or
BED.
In a prospective investigation, Leon et al. (1995) examined precursors to
the later development of an ED in 852 adolescent girls and 815 adolescent
boys. Boys were included to ascertain gender commonalties or differences in

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

654 THE COUNSELING PSYCHOLOGIST / September 2001

the pathways of risk. Students entered the study in grades 7 to 10 and partici-
pated for 3 consecutive years. Variables examined were BMI, pubertal devel-
opment, and personality characteristics. Risk status (high, moderate, mini-
mal, none/control) was determined at each year based on scores on the Eating
Disorder Inventory (Garner, Olmsted, & Polivy, 1983) and an ED checklist of
related behaviors and attitudes. Multiple regression analyses indicated that
for both genders, the strongest predictors of Year 3 risk status were Year 1 and
Year 2 risk scores. When the effects of Year 1 and Year 2 risk were controlled,
ethnicity (White) and poor interoceptive awareness at Year 2 were significant
predictors of disordered eating at Year 3 for girls. Previous risk status was the
only significant predictor of Year 3 risk for boys. The researchers pointed out
that the fact that the strongest predictors of Year 3 risk were Years 1 and 2 risk
scores suggests that when disordered eating becomes a strong pattern, these
habits override individual or group differences in personality, behavior, and
attitudes (Leon et al., 1995). What factors preceded initial risk status at Year 1
for girls or boys remains unclear.
In each of the prospective studies reviewed here, the significant predictor
variable is defined in such as way as to overlap with core symptoms of EDs
(e.g., weight and shape concern and dieting). Hence, the predictor variable is
confounded with the outcome variable of interest, and these studies may be
understood potentially as describing the early course of an ED. This does not
diminish the value of these studies for prevention planning. Clearly, in female
adolescents, weight concerns and dieting increase risk for the exacerbation of
disordered eating over time.
The relationships between obesity and dieting and between obesity and
binge eating appear to be bidirectional. Obesity and dieting have been shown
to predict onset of binge eating in the studies reviewed thus far. However,
dieting and binge eating also have been shown to predict increases in adipos-
ity (Stice & Agras, 1998; Stice, Cameron, et al., 1999). For example, Stice,
Cameron, et al. (1999) examined the prospective relations of weight-reduc-
tion efforts to increases in adiposity and obesity onset in a sample of 692
female adolescents over the course of 4 years. They found that dieting and
binge eating predicted greater weight increase, even when controlling for ini-
tial weight. Perhaps EDs such as BN and BED are the result of a vicious cycle
in which initial efforts to achieve thinness contribute to the onset of disor-
dered eating, which in turn leads to greater weight gain and intensified body
dissatisfaction, prompting an intensification of disordered eating to the point
at which some individuals develop full-syndrome BN or BED.
A recent study examined prospectively the relationship between maternal
and infant characteristics (maternal weight history, maternal eating attitudes
and behaviors, infant birth weight, and infant sucking behavior) and onset of
problem eating in children (Stice, Agras, & Hammer, 1999). Maternal BMI

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

Striegel-Moore, Cachelin / ETIOLOGY OF EDS 655

and maternal disordered eating were predictive of onset of children’s secre-

tive eating and overeating (as reported by their mothers). Moreover, avidity
of sucking during infancy was predictive of overeating. The authors cau-
tioned that disordered eating (e.g., eating in secret, overeating) was transient
in most children. They also noted that for measuring children’s disordered
eating behaviors, future studies need to incorporate instruments of estab-
lished reliability and validity. Nevertheless, this is the first prospective study
of risk factors for eating disturbances in early childhood. Whether these early
problem behaviors are predictive of disordered eating later on needs to be
established.
In conclusion, prospective studies have found support for the role of body
image dissatisfaction and dieting in the development of disordered eating.
Research to date has focused primarily on adolescence, the period when
weight concern is most likely to develop. Given, however, that body image
concern is one of the core symptoms of AN and BN, research needs to follow
younger cohorts (i.e., prepubertal) through adolescence and young adulthood
to identify risk factors for weight concerns. Furthermore, prospective
research needs to extend into young adulthood to identify any risk factors that
are specific to the development of the different categories of EDs (AN, BN, or
BED). Risk factor studies using other designs (e.g., surveys, case-control
studies) have identified several variables that point to high-risk groups (e.g.,
obese girls, girls who have been abused, daughters of mothers with an ED).
By studying high-risk groups longitudinally, risk factor hypotheses can be
tested in a focused way.

IMPLICATIONS FOR FUTURE

RESEARCH AND PRACTICE

Our review suggests several conclusions. Family and twin studies clearly
show that EDs aggregate in families, and research now is needed to elucidate
the mechanisms that underlie this association. Multiple designs will be
needed to explain the significantly elevated risk for EDs among females who
have a relative with an ED. On one hand, genetic research may identify the
genes involved in causing EDs. On the other hand, the phenotypic expression
of a genetic vulnerability is expressed in a particular context. Studies investi-
gating twin pairs who are discordant for an ED may help address the question
of why one twin is protected from developing the disorder even though her
twin sister has developed an ED. This review further illustrates that many of
the personal vulnerability factors have been identified using the case-control
design. Prospective studies are needed to confirm the etiologic significance
of these factors. In addition, for modifiable risk variables, experimental stud-

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

656 THE COUNSELING PSYCHOLOGIST / September 2001

ies such as randomized prevention trials may help clarify the role of personal
vulnerability factors such as low self-esteem or body dissatisfaction. Despite
their great promise, prospective studies have been underutilized, and results
have been limited by the relatively small sample sizes involved in published
studies. The cost involved in carrying out longitudinal studies over a long
time period are considerable, and researchers may need to enlist public sup-
port to generate the interest and resolve needed to commit major resources
toward a better understanding of these disorders.
This review also has identified several major gaps in research. The most
pressing gap is the lack of nationally representative epidemiologic data. A
second, related gap concerns the dearth of data about EDs in ethnic minority
populations. These populations may experience several risk factors that have
not yet been examined with adequate scientific rigor. For example, case
reports describe the acculturation pressures experienced by upwardly mobile
ethnic minority women to adopt a thin body as a way of “fitting in” with
majority culture. A third gap concerns the question of how to define the syn-
dromes of EDs so that they have maximum clinical utility and validity. If risk
factor studies require inclusion of full-syndrome cases (as presently defined),
this poses a serious challenge to certain designs (e.g., prospective cohort
study) because of the costs involved in establishing and following large
cohorts over extended periods of time. Hence, research is needed to deter-
mine how broadly syndromes may be defined to still represent valid variants
of the spectrum of AN, BN, and BED. Moreover, researchers need to move
toward studies in which the entire range of risk domains is measured ade-
quately and the various risk factors can be considered within an integrative
framework. Needed especially are studies that incorporate biological vari-
ables along with other risk factors. Last, the ultimate goal of risk factor
research is to provide data that permit identification of high-risk groups (for
targeted prevention) and development of interventions aimed at eliminating
sources of risk or mediating their impact on those who are exposed. Given the
multitude and complexity of risk factors described in this review, the search
for risk factors needs to be informed by consideration of the distinction of
fixed and modifiable risk factors (Kraemer et al., 1997) and the determination
of the public health benefits of particular preventive interventions versus
early therapeutic interventions.
To date, clinically proven treatments have focused primarily on the per-
sonal vulnerability factors of body image dissatisfaction and dieting (cogni-
tive behavioral therapy) and social self deficits (interpersonal psychother-
apy) (for review, see Whittal, Agras, & Gould, 1999). As noted by Ferguson
and Pigott (2000) and Mitchell (in press), the role of pharmacotherapy
remains limited in the treatment of AN. Moreover, pharmacological treat-

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

Striegel-Moore, Cachelin / ETIOLOGY OF EDS 657

ments of BN and, more recently, BED, are delivered primarily to alleviate

general psychiatric symptoms (e.g., anxiety, depression) and not specifically
on the basis of presumed biological etiological mechanisms.
Experts hope that advances in our understanding of the complex etiologi-
cal mechanisms underlying the development of EDs will contribute to the
development of more efficacious psychological and pharmacological treat-
ments (e.g., Wonderlich, Mitchell, Peterson, & Crow, in press). Further
research may find that constellations of risk factors are associated with cer-
tain prognostic pathways or treatment outcomes. At the individual level,
knowledge of risk factors may be used to tailor treatment to therapeutically
intervene in a more effective manner. At a programmatic level, such informa-
tion can be useful in identifying cost-effective treatment programs and in
identifying high-risk patients early on to mobilize preventive program
resources for such individuals. Finally, the identification of core clusters of
high-risk factors that are collateral with sets of symptoms may be useful in
predicting treatment compliance, which is an important consideration at both
the individual and programmatic levels.

REFERENCES

American Psychiatric Association (1994). Diagnostic and statistical manual of mental disor-
ders (4th ed.). Washington, DC: Author.
Bemis, K. M. (1978). Current approaches to the etiology and treatment of anorexia nervosa. Psy-
chological Bulletin, 85(3), 593-617.
Bordo, S. (1993). Unbearable weight: Feminism, Western culture, and the body. Berkeley: Uni-
versity of California Press.
Boskind-Lodahl, M. (1976). Cinderella’s stepsisters: A feminist perspective on anorexia
nervosa and bulimia. Signs, 2, 342-356.
Bruch, H. (1973). Eating disorders: Obesity, anorexia nervosa and the person within. New York:
Basic Books.
Bruch, H. (1978). The golden cage: The enigma of anorexia nervosa. Cambridge, MA: Harvard
University Press.
Bulik, C. M., Sullivan, P. F., Fear, J., & Pickering, A. (1997). Predictors of the development of
bulimia nervosa in women with anorexia nervosa. The Journal of Nervous and Mental Dis-
ease, 185, 704-707.
Bulik, C. M., Sullivan, P. F., & Kendler, K. S. (1998). Heritability of binge-eating and broadly
defined bulimia nervosa. Biological Psychiatry, 44, 1210-1218.
Bulik, C. M., Sullivan, P. F., Wade, T. D., & Kendler, K. S. (2000). Twin studies of eating disor-
ders: A review. International Journal of Eating Disorders, 27, 1-20.
Bushnell, J. A., Wells, J. E., Hornblow, A. R., Oakley-Browne, M. A., & Joyce, P. (1990). Preva-
lence of three bulimia syndromes in the general population. Psychological Medicine, 20,
671-680.
Cachelin, F. M., & Maher, B. A. (1998). Is amenorrhea a critical criterion for anorexia nervosa.
Journal of Psychosomatic Research, 44, 435-440.

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

658 THE COUNSELING PSYCHOLOGIST / September 2001

Cachelin, F. M., Veisel, C., Striegel-Moore, R. H., & Barzegarnazari, E. (in press). Disordered
eating, acculturation, and treatment seeking in a community sample of Hispanic, Asian,
Black, and White women. Psychology of Women Quarterly.
Chernin, K. (1985). The hungry self: Women, eating, and identity. New York: Harper & Row.
Cnattingius, S., Hultman, C. M., Dahl, M., & Sparen, P. (1999). Very preterm birth, birth trauma,
and the risk of anorexia nervosa among girls. Archives of General Psychiatry, 56, 634-638.
Crago, M., Shisslak, C. M., & Estes, L. S. (1996). Eating disturbances among American minor-
ity groups. International Journal of Eating Disorders, 19, 239-248.
Davis, C., & Katzman, M. A. (1999). Perfection as acculturation: Psychological correlates of
eating problems in Chinese male and female students living in the United States. Interna-
tional Journal of Eating Disorders, 25, 65-70.
Fairburn, C. G., & Beglin, S. J. (1990). Studies of the epidemiology of bulimia nervosa. Ameri-
can Journal of Psychiatry, 4, 402-408.
Fairburn, C. G., Cooper, Z., Doll, H. A., & Welch, S. L. (1999). Risk factors for anorexia nervosa:
Three integrated case-control comparisons. Archives of General Psychiatry, 56, 468-476.
Fairburn, C. G., Cowen, P. J., & Harrison, P. J. (in press). Twin studies and the etiology of eating
disorders. International Journal of Eating Disorders.
Fairburn, C. G., Doll, H. A., Welch, S. L., Hay, P. J., Davies, B. A., & O’Connor, M. E. (1998).
Risk factors for binge eating disorder: A community-based, case-control study. Archives of
General Psychiatry, 55, 425-432.
Fairburn, C. G., Welch, S. L., Doll, H. A., Davies, B. A., & O’Connor, M. E. (1997). Risk factors
for bulimia nervosa. Archives of General Psychology, 54, 509-517.
Fairburn, C. G., Welch, S. L., Norman, P. A., O’Connor, M. E., & Doll, H. A. (1996). Bias and
bulimia nervosa: How typical are clinic cases? American Journal of Psychiatry, 153, 386-
391.
Ferguson C. P., & Pigott, T. A. (2000). Anorexia and bulimia nervosa: Neurobiology and
pharmacotherapy. Behavior Therapy, 31, 237-263.
Furukawa, T. (1994). Weight changes and eating attitudes of Japanese adolescents under
acculturative stresses: A prospective study. International Journal of Eating Disorders, 15,
71-79.
Garfinkel, P. E., & Garner, D. M. (1982). Anorexia nervosa and bulimia. New York: Brunner/
Mazel.
Garfinkel, P. E., Lin, E., Goering, P., Spegg, C., Goldbloom, D. S., Kennedy, S., Kaplan, A. S., &
Woodside, D. B. (1995). Bulimia nervosa in a Canadian community sample: Prevalence and
comparison of subgroups. American Journal of Psychiatry, 152(7), 1052-1058.
Garfinkel, P. E., Lin, E., Goering, P., Spegg, C., Goldbloom, D., Kennedy, S., Kaplan, A. S., &
Woodside, D. B. (1996). Should amenorrhea be necessary for the diagnosis of anorexia
nervosa? British Journal of Psychiatry, 168, 500-506.
Garner, D. M., & Garfinkel, P. E. (1985). Anorexia nervosa: A multidimensional perspective.
New York: Guilford.
Garner, D. M., Olmsted, M. P., & Polivy J. (1983). The development and validation of a multidi-
mensional eating disorder inventory for anorexia nervosa and bulimia. International Journal
of Eating Disorders, 2, 15-34.
Graber, J. A., Brooks-Gunn, J., Paikoff, R. L., & Warren, M. P. (1994). Prediction of eating prob-
lems: An 8-year study of adolescent girls. Developmental Psychology, 30, 823-834.
Gull, W. (1888). Anorexia nervosa. Lancet, 1, 516-517.
Hettema, J. M., Neale, M. C., & Kendler, K. S. (1995). Physical similarity and the equal-
environment assumption in twin studies of psychiatric disorders. Behavior Genetics, 25,
327-335.

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

Striegel-Moore, Cachelin / ETIOLOGY OF EDS 659

Joiner, G. W., & Kashubeck, S. (1996). Acculturation, body image, self-esteem, and eating-dis-
order symptomatology in adolescent Mexican American women. Psychology of Women
Quarterly, 20, 419-435.
Kendler, K. S. (1999). Setting boundaries for psychiatric disorders. American Journal of Psychi-
atry, 156, 1845-1848.
Kendler, K. S., MacLean C., Neale, M., Kessler, R., Heath, A., & Eaves, L. (1991). The genetic
epidemiology of bulimia nervosa. American Journal of Psychiatry, 148, 1627-1637.
Kessler, R. C., Olfson, M.P.H., & Berglund, P. A. (1998). Patterns and predictors of treatment
contact after first onset of psychiatric disorders. American Journal of Psychiatry, 155, 62-69.
Killen, J. D., Taylor, C. B., Hayward, C., Haydel, F. H., Wilson, D., Hammer, L., Kraemer, H.,
Blair-Greiner, A., & Strachowski, D. (1996). Weight concerns influence the development of
eating disorders: A 4-year prospective study. Journal of Consulting and Clinical Psychology,
64, 936-940.
Killen, J. D., Taylor, C. B., Hayward, C., Wilson, D. M., Haydel, K. F., Hammer, L. D.,
Simmonds, B., Robinson, T. N., Litt, I., Varady, A., & Kraemer, H. (1994). Pursuit of thin-
ness and onset of eating disorder symptoms in a community sample of adolescent girls: A
three-year prospective analysis. International Journal of Eating Disorders, 16, 227-238.
Klump, K. L., Holly, A., Iacono, W. G., McGue, M., & Wilson, L. E. (2000). Physical similarity
and twin resemblance for eating disorders: A test of the equal environment assumption.
Behavior Genetics, 30, 51-58.
Kraemer, H. C. (1997). What is the “right” statistical measure of twin concordance (or diagnostic
reliability and validity)? Archives of General Psychiatry, 54, 1121-1124.
Kraemer, H. C., Kazdin, A. E., Offord, D. R., Kessler, R. C., Jensen, P. S., & Kupfer, D. J. (1997).
Coming to terms with the terms of risk. Archives of General Psychiatry, 54, 337-343.
Laseque, C. (1964). De l’anorexie hysterique. [On hysterical anorexia]. In R. M. Kaufman & M.
Heiman (Eds.), Evolutions of psychosomatic concepts: Anorexia nervosa a paradigm. New
York: International University Press. (Original work published 1873)
Leon, G. A., Fulkerson, J. A., Perry, C. L., & Early-Zald, M. D. (1995). Prospective analysis of
personality and behavioral vulnerabilities and gender influences in the later development of
disorder eating. Journal of Abnormal Psychology, 104, 140-149.
Lewinsohn, P. M., Hops, H., Roberts, R. E., Seeley, J. R., & Andrews, J. A. (1993). Adolescent
psychopathology: I. Prevalence and incidence of depression and other DSM-III-R disorders
in high school students. Journal of Abnormal Psychology, 102, 133-144.
Lilenfeld, L. R., & Kaye, W. H. (1998). Genetic studies of anorexia and bulimia nervosa. In H. W.
Hoek, J. L. Treasure, & M. A. Katzman (Eds.), Neurobiology in the treatment of eating disor-
ders (pp. 169-194). New York: Wiley & Sons.
Mitchell, J. (in press). Psychopharmacology of eating disorders: Current knowledge and future
directions. In R. H. Striegel-Moore & L. Smolak (Eds.), Eating disorders: Innovative direc-
tions for research and practice. Washington, DC: American Psychological Association.
Newman, D. L., Moffitt, T. E., Caspi, A., Magdol, L., Silva, P. A., & Stanton, W. R. (1996). Psy-
chiatric disorder in a birth cohort of young adults: Prevalence, comorbidity, clinical signifi-
cance, and new case incidence from ages 11 to 21. Journal of Consulting and Clinical Psy-
chology, 64, 552-562.
Orbach, S. (1986). Hunger strike: The anorectic’s struggle as a metaphor for our age. New
York: Norton.
Patton, G. C., Johnson-Sabine, E., Wood, K., Mann, A. H., & Wakeling, A. (1990). Abnormal
eating attitudes in London school girls: A prospective epidemiological study. Psychiatric
Medicine, 20, 383-394.

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

660 THE COUNSELING PSYCHOLOGIST / September 2001

Regier, D. A., Kaelber, C. T., Rae, D. S., Farmer, M. E., Knauper, B., Kessler, R. C., & Norquist,
G. S. (1998). Limitations of diagnostic criteria and assessment instruments for mental disor-
ders. Archives of General Psychiatry, 55, 109-115.
Robins, L. N., Helzer, J. E., Weissman, M. M., Orvaschel, H., Gruenberg, E., Burke, J. D., &
Regier, D. A. (1984). Lifetime prevalence of specific psychiatric disorders in three sites.
Archives of General Psychiatry, 41, 949-958.
Rodin, J., Silberstein, L. R., & Striegel-Moore, R. H. (1985). Women and weight: A normative
discontent. In T. B. Sonderegger (Ed.), Nebraska symposium on motivation (pp. 267-308).
Lincoln: University of Nebraska Press.
Russell, G. (1979). Bulimia nervosa, an ominous variant of anorexia nervosa. Psychological
Medicine, 9, 429-448.
Shisslak, C. M., Pazda, S. L., & Crago, M. (1990). Body weight and bulimia as discriminators of
psychological characteristics among anorexic, bulimic, and obese women. Journal of
Abnormal Psychology, 99, 380-384.
Smolak, L., & Striegel-Moore, R. H. (in press). Challenging the myth of the golden girl. In R. H.
Striegel-Moore & L. Smolak (Eds.), Eating disorders: Innovative directions for research
and practice. Washington, DC: American Psychological Association.
Stice, E., & Agras, W. S. (1998). Predicting onset and cessation of bulimic behaviors during ado-
lescence: A longitudinal grouping analysis. Behavior Therapy, 29, 257-276.
Stice, E., Agras, W. S., & Hammer, L. D. (1999). Risk factors for the emergence of childhood
eating disturbances: A five-year prospective study. International Journal of Eating Disor-
ders, 25, 375-387.
Stice, E., Cameron, R. P., Killen, J. D., Hayward, C., & Taylor, C. B. (1999). Naturalistic weight-
reduction efforts prospectively predict growth in relative weight and onset of obesity among
female adolescents. Journal of Consulting & Clinical Psychology, 67, 967-974.
Striegel-Moore, R. H. (1993). Etiology of binge eating: A developmental perspective. In C. G.
Fairburn & G. T. Wilson (Eds.), Binge eating (pp. 144-172). New York: Guilford.
Striegel-Moore, R. H. (2000, May 4). Eating disorders 2000: Beyond cultural stereotypes. Paper
presented at the International Conference on Eating Disorders, New York.
Striegel-Moore, R. H., & Cachelin, F. M. (1999). Body image concerns and disordered eating in
adolescent girls: Risk and protective factors. In N. G. Johnson, M. C. Roberts, & J. Worell
(Eds.), Beyond appearance: A new look at adolescent girls (pp. 85-108). Washington, DC:
American Psychological Association.
Striegel-Moore, R. H., Cachelin, F. M., Dohm, F. A., Pike, K. M., Wilfley, D. E., & Fairburn,
C. G. (in press). Comparison of binge eating disorder and bulimia nervosa in a community
sample. International Journal of Eating Disorders.
Striegel-Moore, R. H., & Marcus, M. (1995). Eating disorders in women: Current issues and
debates. In A. L. Stanton & S. G. Gallant (Eds.), Women’s health book (pp. 445-487). Wash-
ington, DC: American Psychological Association.
Striegel-Moore, R. H., Silberstein, L. R., & Rodin, J. (1986). Toward an understanding of risk
factors for bulimia. American Psychologist, 41, 246-263.
Striegel-Moore, R. H., & Smolak, L. (1996). The role of race in the development of eating disor-
ders. In L. Smolak, M. P. Levine, & R. H. Striegel-Moore (Eds.). The developmental
psychopathology of eating disorders (pp. 259-284). Mahwah, NJ: Lawrence Erlbaum.
Striegel-Moore, R. H., & Smolak, L. (2000). The influence of ethnicity on eating disorders in
women. In R. M. Eisler & M. Hersen (Eds.), Handbook of gender, culture, and health
(pp. 227-254). Mawhaw, NJ: Lawrence Erlbaum.
Strober, M. (1995). Disorders of self in anorexia nervosa: An organismic-developmental para-
digm. In C. L. Johnson (Ed.), Psychodynamic treatment of anorexia nervosa and bulimia
(pp. 354-373). New York: Guilford.

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015

Striegel-Moore, Cachelin / ETIOLOGY OF EDS 661

Strober, M., Freeman, R., Lampert, C., Diamond, J., & Kaye, W. (in press). Controlled family
study of anorexia nervosa and bulimia nervosa: Evidence of shared liability and transmission
of partial syndromes. American Journal of Psychiatry.
Sullivan, P. F., Bulik, C. M., & Kendler, K. S. (1998). Genetic epidemiology of binging and vom-
iting. British Journal of Psychiatry, 173 75-79.
Taylor, C. B., Altman, T., Shisslak, C., Bryson, S., Estes, L. S., Gray, N., McKnight, K. M.,
Kraemer, H. C., & Killen, J. D. (1998). Factors associated with weight concerns in adoles-
cents. International Journal of Eating Disorders, 24, 31-42.
van Hoeken, D., Lucas, A. R., & Hoek, H. W. (1998). Epidemiology. In H. W. Hoek, J. L.
Trasure, & M. A. Katzman (Eds.), Neurobiology in the treatment of eating disorders (pp. 97-
126). New York: John Wiley & Sons.
Vitousek, K. B., & Hollon, S. D. (1990). The investigation of schematic content and processing
in eating disorders. Cognitive Therapy and Research, 14, 191-214.
Walker, E. A., Unutzer, J., Rutter, C., Gelfand, A., Saunders, K., VonKorff, A. Koss, M. P., &
Katon, W. (1999). Costs of health care use by women HMO members with a history of child-
hood abuse and neglect. Archives of General Psychiatry, 56, 609-613.
Walters, E. E., & Kendler, K. S. (1995). Anorexia nervosa and anorexic-like syndromes in a pop-
ulation-based female twin sample. American Journal of Psychiatry, 152, 64-71.
Whitaker, A. H. (1992). An epidemiological study of anorectic and bulimic symptoms in adoles-
cent girls. Pediatric Annals, 21, 752-759.
Whittal, M. L., Agras, W. S., & Gould, R. A. (1999). Bulimia nervosa: A meta-analysis of
psychosocial and pharmacological treatments. Behavior Therapy, 30, 117-135.
Wilfley, D. E., Pike, K. M., Dohm, F., Striegel-Moore, R. H., & Fairburn, C. G. (in press). A com-
parison of a community sample and a clinical sample of women with BED. Journal of Con-
sulting and Clinical Psychology.
Wonderlich, S., Mitchell, J., Peterson, C, & Crow, S. (in press). Integrative cognitive therapy for
bulimic behavior. In R. H. Striegel-Moore & L. Smolak (Eds.), Eating disorders: Innovative
directions for research and practice. Washington, DC: American Psychological
Association.
Yanovski, S. Z. (1993). Binge eating disorder: Current knowledge and future directions. Obesity
Research, 1, 306-318.
Zahner, G.E.P., Hsieh, C., & Fleming, J. A. (1995). Introduction to epidemiologic research
methods. In M. T. Tsnang, M. Tohen, & G.E.P. Zahner (Eds.), Textbook in psychiatric epide-
miology (pp. 23-53). New York: Wiley & Sons.

Downloaded from tcp.sagepub.com at MCMASTER UNIV LIBRARY on March 16, 2015