Cogn Ther Res (2012) 36:560–575

DOI 10.1007/s10608-011-9385-8

ORIGINAL ARTICLE

Identification of Child Sexual Abuse Survivor Subgroups Based

on Early Maladaptive Schemas: Implications for Understanding
Differences in Posttraumatic Stress Disorder Symptom Severity
Hilary G. Harding • Erin E. Burns •

Joan L. Jackson

Published online: 10 August 2011

Ó Springer Science+Business Media, LLC 2011

Abstract Cognitive models have informed understanding suggest possible treatment targets for cognitive therapy
of the development, maintenance, and treatment of post- with CSA survivors.
traumatic stress disorder (PTSD). Limited research, how-
ever, has examined the relationship of early maladaptive Keywords Child sexual abuse
Early maladaptive
schemas (EMS; Young in Cognitive therapy for personality schemas (EMS)
Cognitive models
Posttraumatic stress
disorders: A schema-focused approach (rev. ed.). Profes- disorder (PTSD)
sional Resource Press, Sarasota, 1994) to PTSD among
trauma survivors. The current study, using a sample of 127
female child sexual abuse (CSA) survivors, applied a Introduction
model-based clustering procedure (Mclust; Fraley and
Raftery in MCLUST Version 3 for R: Normal Mixture Symptoms of posttraumatic stress, including intrusive and
Modeling and Model-based Clustering, Technical Report recurrent thoughts and images of the traumatic event,
No. 504, Department of Statistics, University of Wash- hyperarousal, and avoidance of trauma-related cues, rep-
ington, 2006) to the 15 subscales of the Young Schema resent common reactions in the aftermath of a traumatic
Questionnaire-Short Form (YSQ-S; Young and Brown in event. Although the majority of people who experience
Young Schema Questionnaire- short form. Cognitive trauma recover in the weeks and months following expo-
Therapy Center, New York, 1994) and revealed three sure, an estimated 5.2 million adults in the U.S. experience
clusters differentiated primarily by level of schema eleva- diagnostic levels of Posttraumatic Stress Disorder (PTSD)
tion. Women in the cluster characterized by the highest in a given year (National Center For PTSD; 2003).
schema scores reported the most severe PTSD symptoms. Available data suggest that women are twice as likely as
A discriminant analysis indicated that schemas of Mistrust/ men to develop PTSD, which may be explained, in part, by
Abuse, Vulnerability to Harm, and Emotional Deprivation evidence that certain types of traumatic events are associ-
contributed most to distinguishing women differentiated on ated with increased rates of posttraumatic stress. For
the basis of presumptive PTSD diagnostic status. Results example, child maltreatment, including child sexual abuse
underscore the importance of cognitive factors in the (CSA), has been associated with increased rates of PTSD
development and/or maintenance of PTSD symptoms and (Breslau et al. 1998; Bromet et al. 1998), with women more
likely to experience CSA than men (Kessler et al. 1994;
Stein and Walker 2000).
Cognitive models have informed our understanding of
Portions of this work were presented at the 2010 World Congress of the etiology and maintenance of PTSD (see reviews by
Behavioral and Cognitive Therapies. Brewin and Holmes 2003; Cahill and Foa 2007; Dalgleish
2004). The majority of cognitive models of PTSD posit
H. G. Harding (&)
E. E. Burns
J. L. Jackson
that negative beliefs about the self, others, and the world
Department of Psychology, University of Georgia, Athens,
GA 30602, USA in the aftermath of a traumatic experience contribute
e-mail: hharding@uga.edu to the development and maintenance of PTSD. Models

123
Cogn Ther Res (2012) 36:560–575 561

that evolved from social-psychological and information- environment is a source of strength (Young et al. 2003).
processing theories invoke the concept of schema and the According to Young et al. (2003), schemas such as
associated processes of assimilation and accommodation to Mistrust/Abuse, Defectiveness/Shame, or Vulnerability to
explain cognitive reactions to traumatic events. Positive Harm are hypothesized to be particularly related to early
pre-trauma schemas may be ineffective in assimilating traumatic or victimization experiences, including exposure
trauma-relevant information (Janoff-Bulman 1992), and to abuse or harm during development. Thus, abusive
thus may be extensively accommodated (or over-accom- experiences during childhood and adolescence may initiate,
modated) in a negative direction. Additionally, pre-existing reinforce, or strengthen EMS.
negative schemas may be strengthened via assimilation of Early maladaptive schemas, like other negative cognitive
trauma-relevant information, as elaborated in several con- styles (e.g., Gibb 2002), have demonstrated associations
temporary models (e.g., Ehlers and Clark 2000; Foa and with childhood adversity and subsequent maladjustment in
Rothbaum 1998; Horowitz 2001; Resick and Schnicke adolescents (Lumley and Harkness 2007) and adults (Harris
1993). As suggested by Brewin and Holmes (2003), there and Curtin 2002; McGinn et al. 2005; O’Dougherty Wright
remains a need for further study of these negative schemas. et al. 2009; Schmidt et al. 1995). Zlotnick and colleagues
In particular, more information is needed about the nature, (1996) found that women with CSA histories exhibited
type, and number of negative schemas relevant to PTSD higher total scores on the Young Schema Questionnaire
and whether certain negative schemas are differentially relative to women without CSA histories. Further, results
relevant to certain types of trauma. from the first study to empirically examine the specificity of
The work of Young and colleagues (2003) regarding Young’s EMS found that schemas relating to themes of
early maladaptive schemas appears particularly relevant to danger (e.g., Mistrust/Abuse and Vulnerability to Harm)
the study of negative pre-trauma schemas and PTSD. These mediated the relationship between general child maltreat-
authors have further articulated the concept of a schema to ment and anxious arousal, while loss/worthlessness schemas
include a ‘‘broad pervasive theme or pattern’’ that is (e.g., Emotional Deprivation, Dependency, Defectiveness,
‘‘comprised of memories, emotions, cognitions, and bodily Failure, and Social Isolation) preferentially mediated the
sensations’’ (Young et al. 2003, p. 7). This conceptualiza- association between maltreatment and anhedonic symptoms
tion proposes that schemas originate in early childhood or (Lumley and Harkness 2007).
adolescence, are both intra- and interpersonal in nature, and To date, there is only one known study that has exam-
become increasingly elaborate with time and experience. ined the relation between EMS and PTSD symptomatology
Young has been most interested in the dysfunctional, self- (Price 2007). In a sample of men and women experiencing
defeating, and impairing schemas that he suspects underlie diagnostic levels of PTSD in response to work-related
psychopathology, and he has consequently labeled them trauma (e.g., events included responding to major transport
early maladaptive schemas (EMS; Young 1994; Young accidents, dealing with serious injury/suicide attempts, or
et al. 2003). being physically attacked), four schemas (Defectiveness,
Researchers have just begun to provide empirical sup- Dependency, Enmeshment, and Failure) significantly pre-
port for Young’s schema constructs. Recent studies suggest dicted PTSD status (Price 2007). It is interesting to note
that schemas are linked to a variety of psychological dis- that the schemas identified by Lumley and Harkness (2007)
orders and associated behaviors including adult depression as predicting anxious arousal did not emerge as significant
(Harris and Curtin 2002; Lumley and Harkness 2007; predictors of PTSD status in the Price (2007) study; rather,
McGinn et al. 2005; O’Dougherty Wright et al. 2009), schemas that related to anhedonic symptoms (i.e., Defec-
anxiety (O’Dougherty Wright et al. 2009), social phobia tiveness, Dependency, and Failure) in Lumley and
(Pinto-Gouveia et al. 2006), eating disorder symptomatol- Harkness (2007) predicted PTSD status in Price (2007). It
ogy (Dingemans et al. 2007; Waller et al. 2000; Waller is possible that child maltreatment samples may demon-
et al. 2007), personality disorders (Petrocelli et al. 2001), strate different associations among EMS and PTSD
self-harm behaviors (Castille et al. 2007), attachment dif- symptomatology relative to the Price (2007) sample of men
ficulties (Mason et al. 2005), interpersonal conflict and women who were exposed to an occupationally-related
(Messman-Moore and Coates 2007) and general psycho- traumatic stressor, due to the potential differences in the
logical distress (Schmidt and Joiner 2004). nature of and context surrounding the traumatic stressor.
Early maladaptive schemas are thought to originate This discrepancy is conceptually interesting and under-
primarily from toxic childhood environments, specifically scores the need for additional research to investigate the
the child’s nuclear family, although they can also result association between EMS and PTSD symptomatology
from negative peer or extra-familial relationships (Young among other trauma-exposed samples.
1994; Young et al. 2003). The latter negative experiences, The present study sought to extend existing research
however, tend to be less powerful especially if the family examining cognitive factors that contribute to the

123
562 Cogn Ther Res (2012) 36:560–575

development or maintenance of PTSD symptoms by Shame) would report more maltreatment in childhood
examining the relation between EMS and PTSD symptoms (Hypothesis Two). Further, it was expected that these
among a sample of young adult female CSA survivors. Child EMS-defined subgroups would demonstrate differences in
sexual abuse survivors would appear to be at risk for the posttraumatic stress symptomatology. Specifically, based
development of EMS due both to their victimization expe- on previous work examining child maltreatment, EMS, and
riences and associated family characteristics. Further, they anxious arousal (e.g., Lumley and Harkness 2007), we
comprise a relatively large group of individuals at risk for expected that subgroups containing individuals with profile
PTSD and other forms of psychopathology. They are also at elevations on Vulnerability to Harm and Mistrust/Abuse
risk for revictimization later in life, such that CSA survivors schemas would demonstrate greater PTSD symptoms
are overrepresented in samples of women being treated for (Hypothesis Three). As an additional validation step, we
PTSD subsequent to adult sexual assault and other traumas. also sought to compare subgroups on additional measures
Therefore, the study of women with CSA histories should of psychopathology (i.e., depressive symptoms and a
provide useful information regarding schemas possibly rel- broader measure of trauma-related affective and behavioral
evant to the development of PTSD symptoms. No study to symptomatology), to further examine the explanatory
date, however, has examined the relationship of EMS to capacity of the EMS-defined subgroups.
PTSD in a sample of child sexual abuse survivors. Finally, we also applied a more traditional, variable-
We chose to examine these relationships initially using a oriented data analytic strategy (Nurius and Macy 2008) to
person-centered analytic approach (Nurius and Macy determine which early maladaptive schemas were most
2008), given the variability typically observed among CSA predictive of PTSD symptomatology in the sample of CSA
survivors in regard to abuse characteristics, family of origin survivors as a whole. It was expected that Vulnerability to
characteristics, co-occurring forms of maltreatment, and Harm and Mistrust/Abuse schemas would emerge as sig-
psychological outcomes. Person-centered approaches to nificant predictors, given their theoretical connection to
data analysis, in contrast to variable-centered approaches, early abuse or trauma history (Young et al. 2003) and their
focus on persons rather than variables. A primary goal is to empirical association to anxious arousal symptomatology
identify meaningful subgroups of persons whose charac- (Lumley and Harkness 2007) (Hypothesis Four).
teristics might be obscured when only group averages are
considered. When applied in clinical populations, the
objective is often to identify subgroups of persons who Method
might be responsive to different interventions. Person-
centered research methods have been seen as particularly Participants
applicable in studying violence-exposed women due to the
considerable heterogeneity within this population (Nurius Participants comprised a subset of a larger sample of
and Macy 2008). The present study used a cluster analytic undergraduate women previously recruited for a study of
strategy to identify subgroups of CSA survivors on the child maltreatment. Women in the present study included
basis of early maladaptive schema subscale scores. those who endorsed experiencing child sexual abuse (CSA)
The current hypotheses were developed based on the (n = 127) and a comparison group of women who did not
theoretical and empirical literature supporting the associ- report CSA (n = 50). Given that previous research has
ation between childhood maltreatment and EMS (e.g., established racial/ethnic differences in self- reported CSA
Young et al. 2003; Zlotnick et al. 1996), and the empirical experiences among college women in regard to reported
work most conceptually relevant to the current study—the prevalence, severity, victim-offender relationship, and
work of Lumley and Harkness (2007) that has demon- post-abuse coping strategies (Ullman and Filipas 2005), we
strated an association between child maltreatment, specific chose to select a comparison sample of non-CSA women
abuse-related EMS (i.e., Vulnerability to Harm and Mis- who were similar to the CSA subsample on the basis of
trust/Abuse) and anxious arousal symptomatology. First, it racial/ethnic background in order to control for potential
was expected that different subgroups of survivors would confounds presented by demographic differences in the
be present in our sample, with subgroups displaying dif- CSA and non-CSA groups. Thus, a stratified random
ferent profiles, or patterns, of schema elevations (Hypoth- selection procedure was used in order to select a non-CSA
esis One). Next, it was expected that the validity of the sample that reflected the same proportion of women from
EMS-defined subgroups would be supported by group diverse racial/ethnic backgrounds as was present in the
differences in childhood maltreatment experiences. Spe- CSA sample. Since we hypothesized the presence of sub-
cifically, it was hypothesized that subgroups characterized groups in our CSA sample, fifty women were selected for
by the presence of high levels of abuse-related EMS (i.e., the non-CSA subgroup in order to balance the sample size
Vulnerability to Harm, Mistrust/Abuse, and Defectiveness/ of each group for purposes of the analyses.

123
Cogn Ther Res (2012) 36:560–575 563

The mean age of the CSA survivor subsample was subscale was dropped from subsequent analyses due to low
19.3 years (SD = 2.54), and that of the non-CSA sub- internal consistency.
sample was 19.0 years (SD = 1.11). No significant dif- The sexual abuse subscale of the CTQ was used to
ferences were observed across the CSA and non-CSA identify the current sample of female survivors of child
comparison group in regard to age, t(165) = .75, p = .20. sexual abuse. Specifically, this study used an established
With regard to racial/ethnic background, the combined clinical cutoff point for CSA (i.e., a score of 6 or above on
sample self-identified as White/Caucasian (58.2%), Black/ the CSA subscale) to determine the presence of abuse. The
African–American (18.1%), Asian or Pacific Islander scale developers established this clinical cutoff score using
(14.1%), American Indian or Alaska Native (.6%), and two large independent samples, including a community
Hispanic or Latina (2.8%) (6.2% did not indicate their sample of 1,225 women and a sample of 378 treatment-
racial/ethnic identification). A Chi-Square test was con- seeking substance abusers. They compared scores on the
ducted to compare frequencies of White and non-White CTQ to clinician-rated data as the criterion for computing
participants in the CSA and non-CSA subgroups. As sensitivity and specificity (Bernstein and Fink 1998). With
expected based on our selection procedure for the non-CSA a clinical cutoff point of 6, the CSA subscale identified
comparison group, there were no significant differences child sexual abuse cases based on clinician ratings with
observed in the racial/ethnic composition of the two 83% sensitivity and 83% specificity in a sample of
groups, v2(df = 3) = 1.68, p = .64. 151 female health maintenance organization members
(Bernstein and Fink 1998).
Procedure
Early Maladaptive Schemas
Participants were recruited through a research participant
(RP) pool and received credit toward course requirements The Young Schema Questionnaire-Short Form (YSQ-S;
for participation in the study. Survey and procedures were Young and Brown 1994) contains 75 items rated on a
approved by the university’s IRB. Participants completed 6-point scale, ranging from 1 (completely untrue of me) to
paper-and-pencil, self-report measures of demographic 6, (describes me perfectly). The YSQ-S has well-
information, child maltreatment history, and posttraumatic established subscale internal consistency (Schmidt et al.
stress symptomatology. 1995; Waller et al. 2001). Young (1999) proposed that each
item belongs to one of 15 theoretically-derived early
Measures maladaptive schema subscales that comprise five higher-
order domains. The Disconnection and Rejection domain is
Childhood Maltreatment comprised of schemas relating to themes of Abandonment
(‘‘I find myself clinging to people I’m close to, because I’m
The Childhood Trauma Questionnaire (CTQ; Bernstein and afraid they’ll leave me’’), Mistrust/Abuse (‘‘I am quite
Fink 1998) was used to assess childhood sexual abuse and suspicious of other people’s motives’’), Emotional Depri-
other forms of childhood maltreatment. The CTQ is a brief vation (‘‘For much of my life, I haven’t felt that I am
self-report inventory inquiring about childhood sexual special to someone’’), and Defectiveness/Shame (‘‘I am too
abuse (CSA), childhood emotional abuse (CEA) and unacceptable in very basic ways to reveal myself to other
neglect (CEN), and childhood physical abuse (CPA) and people’’). The Impaired Autonomy domain is comprised of
neglect (CPN) experiences in childhood and adolescence. schemas relating to Social Isolation (‘‘I don’t fit in’’),
Items are responded to on a 5-point Likert scale, with Dependence (‘‘I do not feel capable of getting by on my
response options ranging from 1 (Never True) to 5 (Very own in everyday life’’), Vulnerability to Harm (‘‘I worry
Often True). Validation of the CTQ has involved multiple about being attacked’’), Enmeshment (‘‘I often feel that I do
clinical and community samples (e.g., Bernstein and Fink not have a separate identity from my parent or partner’’),
1998; Fink et al. 1995). Studies indicate that this scale has and Failure (‘‘I’m incompetent when it comes to
high temporal stability (evidenced by high test–retest achievement’’). The Impaired Limits domain is comprised
reliability over a 2- to 6-month interval; intraclass corre- of schemas relating to Entitlement (‘‘I hate to be con-
lation = .88), high internal consistency (Cronbach’s alphas strained or kept from doing what I want to do’’) and
ranging from .79 to .94) and good convergence with Insufficient Self-Control (‘‘I can’t seem to discipline
standardized interviews (Bernstein and Fink 1998). Internal myself to complete routine or boring tasks’’). The Other-
consistency for the CSA, CPA, CEA, CEN, and CPN Directedness domain is comprised of schemas relating to
subscales within the current sample were moderate to high, Subjugation (‘‘In relationships, I let the other person have
with the exception of the CPN subscale, Cronbach’s the upper hand’’) and Self- Sacrifice (‘‘Other people see me
as = .90, .68, .86, .97, and .54, respectively. The CPN as doing too much for others and not enough for myself’’);

123
564 Cogn Ther Res (2012) 36:560–575

and finally, the Overvigilance and Inhibition domain is adult interpersonal assault histories scoring higher, as
comprised of schemas relating to Emotional Inhibition predicted, than individuals who experienced natural
(‘‘I control myself so much that people think that I am disasters and accidents (Lauterbach and Vrana 1996;
unemotional’’) and Unrelenting Standards (‘‘I must Zinzow and Jackson 2009). Cronbach’s alpha for the total
meet all my responsibilities’’). scale score in the present sample was .95 (see Table 1).
The first psychometric evaluation of the YSQ using No diagnostic cut-off has been established for the
samples of undergraduate students and adults from the PPTSD-R total score. However, a small body of normative
community demonstrated adequate test–retest reliability data on the scale is available that can provide a useful
(coefficients ranging from .50 to .82) and internal consis- reference point for the present study. Lauterbach and Vrana
tency (alpha coefficients ranging from .83 to .96) (Schmidt (1996) found that college women reporting histories of
et al. 1995). Recent research suggests that the YSQ-S and child abuse (broadly defined) (N = 11) had a mean total
the original, longer version (which has been more widely score of 36.0 (SD = 12.8), whereas women with histories
evaluated) demonstrate similar levels of internal consis- of rape had mean scores of 40.0 (SD = 15.4). Similarly,
tency, discriminant validity, and clinical utility (Waller Zinzow and Jackson (2009) found that college women with
et al. 2001). Given the reduced administration time and histories of child abuse (N = 75) had a mean score of 36.2.
comparable psychometric properties, the YSQ-S was used Clinical samples, as expected, evidence somewhat higher
in the current study. Alpha coefficients for each of the average scores. Women in a clinical sample of both trauma
subscales ranged from .61 to .94, and are presented in exposed and non-exposed individuals had a mean score of
Table 1. 41.4 (SD = 14.8) (Lauterbach and Vrana 1996). In a
sample of 14 mothers of abused children, those who
Posttraumatic Stress Symptoms themselves had histories of CSA had a mean PPTSD-R
score of 45.0 (SD = 13.6) (Timmons-Mitchell et al. 1996).
The Purdue PTSD Scale-Revised (PPTSD-R; Lauterbach
and Vrana 1996) is a 17-item self-report scale that assesses Depressive Symptoms
the symptoms comprising DSM-IV (American Psychiatric
Association 1994) criteria B, C, and D for PTSD. Past The Beck Depression Inventory-II (BDI-II; Beck et al.
month symptoms are rated on a scale from 1 (not at all) to 5 1996) was used to assess participants’ depressive symp-
(often), with a midpoint of 3 (sometimes). Total scores can toms. The BDI-II consists of 21 items assessing the cog-
range from 17 to 85, with higher scores reflecting more nitive and affective/somatic symptoms of depression, and
severe symptomatology. In addition to measuring symptom its psychometric properties have been well established
severity via the total score, the PPTSD-R can also be used among both adult (Beck et al. 1996) and college student
to provide a presumptive diagnosis via application of DSM samples (Steer and Clark 1997). Items are scored from 0 to
threshold criteria (Lauterbach and Vrana 1996). This 3, with higher scores reflective of greater depressive
measure was also used in the present study. Individuals symptomatology. Total scores on this measure can be
were considered PTSD positive if they endorsed one or interpreted as follows: raw scores ranging from 0 to 13 are
more re-experiencing symptoms, three or more avoidance suggestive of minimal depression; 14–19 suggestive of
symptoms, and two or more arousal symptoms occurring at mild depression; 20–28 suggestive of moderate depression;
least ‘‘sometimes’’ (i.e., a rating of 3 or higher on the 1–5 and scores ranging from 29 to 63 are indicative of severe
scale). levels of depressive symptoms (Beck et al. 1996). Cron-
The PPTSD-R displayed adequate psychometric prop- bach’s alpha of the total score in the current sample was .93
erties in a series of three studies conducted by Lauterbach (see Table 1).
and Vrana (1996). Both adequate test–retest reliability
(r = .71 over 2-weeks) and internal consistency of the total Trauma Symptoms
score (Cronbach’s alpha = .91) were reported. Conver-
gent validity was demonstrated via moderate to high cor- Psychological sequelae of childhood sexual abuse fre-
relations with other measures of posttraumatic stress quently include disturbances beyond the PTSD-specific
symptomatology and discriminant validity by smaller, hyperarousal, reexperiencing, and avoidance/numbing
moderate-sized correlations with depression and trait anx- symptom clusters. Thus, we included the Trauma Symptom
iety. The PPTSD-R also discriminated individuals seeking Inventory (TSI; Briere 1995) to assess participants’ expe-
treatment for trauma-related problems from individuals riences of a broad range of trauma-related psychological
with other presenting problems. Further, scores on the distress. The TSI is a 100-item inventory with items
PPTSD-R discriminated groups based on type of traumatic assessing affective disturbances (i.e., depression, anger/
event, with individuals with child abuse, rape and other irritability), PTSD-specific symptoms (i.e., anxious arousal,

123
 Table 1 Means, standard deviations, reliability coefficients, and correlations for measures of childhood maltreatment, PTSD symptoms, depressive symptoms, trauma symptoms, and Early
Maladaptive Schemas for the entire sample
Variable M SD 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.

1. Child sexual abuse 8.9 4.9 .90 .24 .35 .33 .28 .17 .27 .20 .34 .26 .15 .21 .16 .19 .03 .09 .12 -.01 .11 .21 .16 .20
Cogn Ther Res (2012) 36:560–575

2. Child physical abuse 6.9 2.6 .68 .71 .56 .27 .22 .30 .28 .32 .42 .19 .19 .20 .26 .16 .12 .05 .15 .18 .12 .24 .08
3. Child emotional abuse 8.9 4.5 .86 .78 .39 .43 .49 .43 .45 .59 .35 .33 .44 .40 .18 .21 .05 .24 .30 .24 .39 .07
4. Child emotional neglect 8.4 4.1 .97 .24 .32 .32 .42 .37 .62 .27 .30 .39 .36 -.01 .21 -.06 .23 .22 .18 .32 -.10
5. PTSD symptoms 33.0 15.9 .95 .57 .75 .48 .53 .37 .43 .37 .32 .52 .28 .31 .19 .21 .39 .27 .43 .29
6. Depressive symptoms 12.1 10.3 .93 .74 .56 .54 .40 .56 .47 .41 .62 .29 .51 .28 .44 .47 .24 .54 .22
7. TSI total score 70.5 38.5 .97 .54 .50 .43 .45 .42 .35 .59 .24 .32 .23 .37 .38 .22 .46 .20
8. Abandonment 10.7 5.9 .89 .60 .55 .55 .44 .41 .54 .30 .45 .26 .41 .55 .28 .52 .13
9. Mistrust/Abuse 11.7 5.9 .90 .62 .56 .56 .33 .60 .34 .40 .41 .40 .47 .32 .61 .30
10. Emotional deprivation 9.4 5.3 .87 .47 .49 .43 .53 .31 .37 .21 .28 .35 .26 .55 .11
11. Defectiveness 8.0 5.4 .92 .72 .47 .43 .27 .63 .11 .39 .53 .31 .66 .26
12. Social isolation 10.1 6.1 .93 .37 .43 .31 .40 .20 .40 .54 .28 .54 .25
13. Dependence 7.8 3.2 .61 .30 .23 .57 .09 .33 .44 .24 .37 .00
14. Vulnerability 8.1 4.1 .82 .31 .33 .29 .39 .45 .27 .54 .26
15. Enmeshment 7.6 3.8 .82 .25 .40 .15 .46 .01 .29 .26
16. Failure 9.0 5.5 .94 .11 .45 .44 .16 .42 .05
17. Entitlement 11.4 5.2 .81 .46 .20 -.01 .26 .31
18. Self-control 10.5 5.7 .90 .40 .04 .36 .08
19. Subjugation 9.4 5.0 .84 .39 .49 .27
10. Self-sacrifice 15.4 5.6 .84 .26 .43
21. Emotional inhibition 10.1 5.7 .85 .34
22. Unrelenting standards 18.1 6.4 .88
N = 157 using list-wise deletion of missing data. Bolded correlations are significant at p B .001, two-tailed. Underlined correlations are significant at p B .01, two-tailed. Alpha reliability
coefficients are presented in italics on the diagonal
TSI trauma symptom inventory
565

123
566 Cogn Ther Res (2012) 36:560–575

intrusive experiences and defensive avoidance), and per cluster (see Lo et al. 2001; Ning and Finch 2004; as
behavioral and cognitive disturbances that are frequently cited in Hicks et al. 2004). Moreover, a brief literature
associated with sexual abuse and victimization (i.e., dis- review of ten published studies using Mclust revealed
sociation, sexual concerns, dysfunctional sexual behavior, sample sizes ranging from 36 to 312 (mean sample
impaired self-reference, and tension-reduction behaviors) size = 117). Thus, our sample size of 127 participants is
(Briere 1995). The TSI has strong psychometric properties, consistent with existing research using this methodology
demonstrated by validation studies that have established and appears to exceed a recommended lower-end sample
convergent, discriminant, and criterion validity for this size appropriate for this statistical approach.
instrument, as well as good internal consistency of each Following the cluster analysis, a series of one-way
subscale (Briere 1995). The total scale score was used in ANOVAs were conducted to examine group mean differ-
the current study, and Cronbach’s alpha of the total scale ences on each of the schema subscale scores. Alpha was
score was .97 (see Table 1). adjusted according to the number of analyses conducted in
order to control for Type I error. It was expected that
Design and Analyses clusters would differ significantly on the criterion variables
used to create the clusters (i.e., the schema subscale
First, a model-based cluster analysis using the Mclust scores). Groups were also compared on measures of child
program developed for R software (Fraley and Raftery maltreatment, including CSA as well as physical abuse and
1998) was carried out on the EMS data to identify different emotional abuse and neglect. Next, groups were compared
schema profiles among CSA survivors. Model-based clus- on PTSD symptom severity, depressive symptom severity,
ter analysis overcomes the primary limitations of more and trauma-related difficulties using a series of one-way
commonly used clustering techniques; specifically, the ANOVAs. Finally, a discriminant analysis was conducted
tendency for cluster techniques that specify different to examine which schema subscales were most efficient in
cluster characteristics to produce different cluster classifi- differentiating women with and without a presumptive
cations using the same data, the imposition of a cluster PTSD diagnosis.
solution on data that may not actually have a cluster
structure, and the inability to determine goodness-of-fit for
cluster solutions (Mun et al. 2008). Model-based cluster Results
analysis tests the fit of models that differ in the specifica-
tion of cluster characteristics, and account for data char- Model-Based Cluster Analysis: Cluster Results
acteristics including: (1) clusters that differ in orientation,
size, or shape; (2) non-normally distributed clusters, and The best fitting model (BIC value = -4,881) yielded a
(3) ‘‘noise’’ in the data (Banfield and Raftery 1993). three-cluster solution with diagonal clusters that varied in
Model-based cluster analysis is a relatively new proce- volume and shape. The next best fitting model (BIC
dure that has been used to examine a number of psycho- value = -4,917) also yielded a three-cluster solution but
logical constructs in recent years, including adolescent and with components of differing shape, orientation, and vol-
young adult delinquency, smoking, substance use, sexual ume. When comparing models, a difference in BIC of
behaviors (e.g., Mun et al. 2008), subtypes of psychopathy greater than 10 is considered very strong support for the
(e.g., Hicks et al. 2004; Lee et al. 2010; Skeem et al. 2007), better fitting model (Raftery 1995). Additionally, the mean
and sensory processing subtypes in children with Autistic posterior probability for cluster membership in the best-
Disorder (Lane et al. 2010). In the current study, stan- fitting solution was .98. Moreover, 93% of cases (n = 114)
dardized scores on each of the 15 schema subscales were were classified with probabilities greater than .95. Taken
entered into the analysis. The number of clusters was together, there was substantial evidence that the identified
determined by the best fitting model identified by the three-cluster solution was a good fit to the structure of the
clustering procedure, as defined by the Bayesian Informa- data, and thus the first three-cluster solution was selected
tion Criterion (BIC). The average posterior probability of and interpreted as the optimal model. The age and racial/
cluster membership was also computed to indicate the level ethnic background of the women in each CSA cluster and
of classification certainty for the cases in our sample based the non-CSA comparison group are presented in Table 2.
on the identified cluster solution.
Researchers have noted that simulation studies using Cluster Differences: Early Maladaptive Schemas
Mclust and mixture modeling suggest that this methodol-
ogy can be used with relatively small sample sizes (i.e., Results of the cluster analysis revealed that, in general,
around 50 cases) provided that the identified clusters have CSA subgroups were differentiated predominantly by the
distinct means and have a similar number of observations severity of schema subscale scores. Schema profiles for

123
Cogn Ther Res (2012) 36:560–575 567

Table 2 Age and racial/ethnic identification of participants across CSA subgroups and non-CSA comparison group
Low-EMS (n = 44) Mod-EMS (n = 59) High-EMS Non-CSA (n = 50) F(3,165) v2(df = 3)a p
(n = 20)
M(SD) n % M(SD) n % M(SD) n % M(SD) n %

Participant age 19.2(1.5) 19.7(3.4) 18.6(.8) 19.0(2.3) 1.60 .19

Race/Ethnicity
White/Caucasian 27 61.4 38 64.4 10 50.0 29 58.0 1.68 .64
Black/African– 9 20.5 9 15.3 5 25.0 8 16.0
American
Asian or Pacific 6 13.6 7 11.9 3 15.0 7 14.0
Islander
Hispanic/Latina 1 2.3 3 5.1 0 .0 1 2.0
Did not respond 1 2.3 1 1.7 2 10.0 5 10.0
a
A Chi-square test was conducted to compare frequencies of White vs. Non-White participants across CSA subgroups and the non-CSA
comparison group

each of the three clusters are depicted graphically in Fig. 1. schema elevations (as opposed to different patterns of
Table 3 displays the means, standard deviations, and schema elevations).
results of omnibus ANOVAs testing group differences on
each of the schema subscales. Least Significant Differences Cluster Validation: Cluster Differences in CSA
(LSD) post hoc tests examined subgroup-level differences Severity and Other Forms of Maltreatment
on schema subscales. CSA subgroups differed significantly
on each of the schema subscales. Four cases (3%) were not Next, we sought to examine variables that might further
classified due to missing data in the schema subscales. The validate the observed subgroups, by comparing subgroups
first cluster, Low-EMS (n = 44, 35.8% of the sample) on measures of CSA and other forms of child maltreat-
exhibited the lowest scores on each of the schema sub- ment. Subgroups of CSA survivors and the non-CSA
scales. The next cluster, Moderate-EMS (n = 59, 48.0% of comparison group were compared on retrospective reports
the sample) exhibited mid-level scores on each of the of childhood sexual abuse, childhood emotional abuse and
schema subscales. A third cluster, High-EMS (n = 20, neglect and childhood physical abuse. Recall that we
16.3% of the sample) emerged as the subgroup with the expected that the subgroup membership based on early
highest elevations on each of the schema subscales. maladaptive schemas would reflect differences in regard to
Notably, this cluster demonstrated particularly high ele- childhood maltreatment experiences, such that subgroups
vations on the subscales measuring Abandonment, Mis- characterized by the presence of high levels of abuse-
trust/Abuse, Defectiveness, Social Isolation, Subjugation, related EMS (i.e., Vulnerability to Harm, Mistrust/Abuse,
and Emotional Inhibition (as evidenced by mean scores and Defectiveness/Shame) would report more maltreat-
greater than 1.5 standard deviations above the CSA sample ment in childhood. Based on the subgroups identified with
mean, see Fig. 1). Not surprisingly, results of ANOVAs the cluster analysis, this meant that we expected the High-
revealed that the greatest effect sizes of subgroup mem- EMS cluster to report the greatest levels of maltreatment.
bership were also found for each of these subscales (g2s Results of a series of one-way ANOVAs are presented in
ranging from = .40 to .48, see Table 3). Overall, the Table 4, illustrating significant effects of group member-
largest effect sizes for subgroup membership were found ship for each of these variables, with omnibus effect sizes
for the five schemas within the Disconnection/Rejection ranging from .19 to .33. Least Significant Differences
domain and for the Subjugation and Emotional Inhibition (LSD) post hoc tests examined group-level contrasts in
schemas. Schemas in the Impaired Autonomy and Impaired regard to these variables.
Limits domains exhibited smaller, more moderate effects of Results indicated that the Low-EMS, Moderate-EMS,
subgroup membership (see Table 3). and High-EMS subgroups did not differ significantly in
In summary, our first hypothesis, that there would be regard to the level of sexual abuse sustained in childhood
evidence of CSA subgroups that differed in regard to their (all three groups, as expected, reported significantly more
EMS profile patterns, was only partially supported. Instead, CSA than the non-CSA comparison group). However, the
our results suggested that the primary feature differentiat- subgroups did differ in regard to childhood physical
ing the CSA subgroups identified was the severity of abuse and childhood emotional abuse and neglect. The

123
568 Cogn Ther Res (2012) 36:560–575

Fig. 1 Subgroups of female child sexual abuse (CSA) survivor Emotional Deprivation; Def = Defectiveness; S.I. = Social Isolation;
subgroups defined by a model-based cluster analysis of standardized Dep = Dependence; Vuln = Vulnerability to Harm; Enmesh =
scores on a measure of Young’s Early Maladaptive Schemas (EMS). Enmeshment; Fail = Failure; Ent. = Entitlement; I.S.C. = Impaired
Cluster 1 = Low-EMS; Cluster 2 = Moderate-EMS; Cluster 3 = Self Control; Sub. = Subjugation; S–S = Self-Sacrifice; E.I. = Emo-
High-EMS; Ab = Abandonment; M/A = Mistrust/Abuse; E.D. = tional Inhibition; U.S. = Unrelenting Standards

Table 3 Child sexual abuse subgroups and non-CSA comparison group means on 15 Early Maladaptive Schema subscales
Low-EMS Mod-EMS High-EMS Non-CSA F p g2
(n = 44), M (SD) (n = 59), M (SD) (n = 20), M (SD) (n = 50), M (SD) (3, 172)

Domain: disconnection
Abandonment 7.6 (3.5)a 12.1 (4.5)b 19.7 (7.0)c 8.2 (3.6)a 41.90 \.001 .43
a b c a
Mistrust/Abuse 8.4 (2.8) 13.3 (4.4) 21.4 (6.7) 9.4 (4.2) 49.00 \.001 .47
Emotional deprivation 6.3 (2.3)a 11.4 (4.9)b 16.1 (5.9)c 7.1 (3.5)a 34.92 \.001 .38
Defectiveness 5.1 (.3)a 8.6 (3.3)b 16.6 (8.2)c 6.3 (3.8)a 43.30 \.001 .44
Social isolation 6.8 (2.0)a 11.3 (4.9)b 19.1 (7.1)c 8.1 (4.7)a 37.20 \.001 .40
Domain: impaired autonomy
Dependence 6.4 (1.8)a 8.6 (3.1)b 10.8 (4.5)c 6.7 (2.4)a 15.11 \.001 .21
a b c
Vulnerability 6.3 (1.9) 9.5 (3.5) 13.0 (6.2) 6.5 (2.8)a 24.48 \.001 .30
Enmeshment 6.4 (1.9)a 7.9 (3.2)b 12.6 (7.3)c 6.7 (2.6)a,b 16.24 \.001 .22
Failure 5.5 (1.2)a 10.5 (4.7)b 14.9 (7.7)c 8.0 (4.6)d 22.20 \.001 .28
Domain: impaired limits
Entitlement 9.7 (4.1)a 12.6 (4.6)b 14.6 (7.9)b 10.6 (4.3)a 6.01 .001 .10
a b c
Self-control 7.6 (2.8) 11.9 (5.0) 16.3 (8.1) 9.8 (5.1)d 15.45 \.001 .22
Domain: other-directedness
Subjugation 6.7 (2.6)a 9.8 (3.9)b 18.4 (6.8)c 7.9 (2.5)a 49.44 \.001 .47
a b c
Self-sacrifice 13.4 (5.1) 16.4 (5.2) 19.9 (6.6) 14.7 (5.0)a,b 8.01 \.001 .13
Domain: over-vigilance and inhibition
Emotional inhibition 6.9 (2.5)a 11.0 (4.7)b 19.4 (5.8)c 7.9 (3.6)a 48.15 \.001 .46
a a b
Unrelenting standards 17.1(6.2) 18.0 (6.2) 23.7 (6.1) 16.9 (5.5)a 6.95 \.001 .11
Means with a common superscript do not significantly differ, ps [ .05. Bonferonni corrected alpha was set to .003

High-EMS group endorsed the greatest level of childhood intermediate level of CPA. In regard to emotional neglect,
physical abuse, a level that was significantly greater than the the Moderate-EMS and High-EMS subgroups endorsed the
level reported by the Low-EMS groups and the non-CSA highest levels of emotionally neglectful experiences in
comparison group. The Moderate-EMS group endorsed an childhood, and were significantly greater than the levels

123
Cogn Ther Res (2012) 36:560–575 569

Table 4 Child sexual abuse subgroups and non-CSA comparison group means on measures of childhood maltreatment and psychopathology
Low-EMS Mod-EMS High-EMS Non-CSA F df p g2
(n = 44), M (SD) (n = 59), M (SD) (n = 20), M (SD) (n = 50), M (SD)

Group differences: childhood maltreatment

Child sexual abuse 9.4 (4.7)a 10.4 (5.0)a 11.5 (5.8)a 5.0 (.0)b 18.26 3,170 \.001 .26
a a,b b
Child physical abuse 6.9 (2.4) 7.7 (2.7) 8.6 (3.7) 5.4 (.7)c 12.38 3,169 \.001 .19
Child emotional abuse 8.1 (3.7)a 10.4 (4.5)b 13.9 (5.6)c 5.8 (1.1)d 25.50 3,166 \.001 .33
Child emotional neglect 7.4 (3.0)a 9.9 (4.6)b 11.7 (5.6)b 6.1 (1.4)a 15.42 3,168 \.001 .23
Group differences: psychopathology
PTSD symptoms 29.1 (11.2)a 35.8 (15.3)b 52.0 (14.8)c 26.0 (13.4)a 18.29 3,164 \.001 .25
Depressive symptoms 9.0 (7.7)a 13.7 (8.6)b 25.0 (12.0)c 7.3 (7.9)a 21.39 3,164 \.001 .29
TSI: total scoree 60.3 (29.9)a 78.1 (35.0)b 113.3 (44.0)c 53.5 (30.1)a 15.97 3,157 \.001 .24
Means with a common superscript do not significantly differ, ps [ .05. For omnibus tests, Bonferonni-corrected alpha was set to .007
TSI = Trauma Symptom Inventory
e
Group comparisons were also conducted on the individual subscales of the TSI (i.e., Anxious Arousal, Depression, Intrusive Experiences,
Defensive Avoidance, Dissociation, Anger/Irritability, Dysfunctional Sexual Behavior, Impaired Self-Reference, Sexual Concerns, and Tension-
Reduction Behavior). All subscales followed the same pattern of means across subgroups, with the greatest elevations found in the High-EMS
subgroup, intermediate elevations displayed by the Moderate-EMS subgroups, and lowest mean levels displayed by the Low-EMS and Non-CSA
groups

reported by the Low-EMS subgroup and the non-CSA PTSD symptoms (M = 35.8, SD = 15.26), a level that was
comparison group. In regard to emotional abuse, the High- significantly more severe than both the Low-EMS subgroup
EMS group endorsed the highest level of CEA, and was (M = 29.1, SD = 11.24) and the non-CSA comparison
significantly different from all other EMS subgroups and group (M = 26.0, SD = 13.37), ps \ .05. The Low-EMS
the non-CSA comparison group. To summarize, the CSA subgroup and the non-CSA comparison group did not differ
subgroups in general reported more abuse relative to the in regard to mean level of PTSD symptom severity (see
non-CSA comparison group, and the Moderate- and High- Table 4). By way of interpretation of these scores, it can be
EMS subgroups appear to comprise the portion of the noted that the mean score of the Moderate-EMS subgroup
sample reporting the greatest levels of childhood maltreat- closely corresponds to the mean scores previously found in
ment. Taken together, results indicate a general pattern college women reporting histories of CSA (Lauterbach and
consistent with our hypothesis that the EMS subgroup Vrana 1996; Zinzow and Jackson 2009). Further, the mean
characterized by the presence of higher levels of abuse- score of the High-EMS subgroup slightly exceeds that
related schemas (i.e., the High-EMS group) would endorse found in a clinical sample of CSA survivors.
greater levels of childhood maltreatment (See Table 4). Presumptive diagnoses derived from the PPTSD-R data as
previously described also support subgroup differences in
Cluster Differences: PTSD Symptoms terms of prevalence, v2(3, N = 161) = 39.53, p \ .001. Of
women in the High-EMS subgroup, 80% were PTSD posi-
Next, subgroups of CSA survivors and the non-CSA com- tive. In the Moderate-EMS subgroup, 33% were PTSD
parison group were compared on levels of PTSD symp- positive. Only 7% of women in the Low-EMS were PTSD
tomatology. It was expected that the group with the highest positive, and 17% of those in the comparison group who
schema elevations on Vulnerability to Harm and Mistrust/ completed the PPTSD-R were PTSD positive. (In the com-
Abuse (i.e., the High-EMS cluster) would also exhibit the parison group only participants who reported a potentially
greatest mean levels of PTSD symptoms. Results of a one- traumatic event (other than CSA) completed the scale.)
way ANOVA revealed a significant overall effect of group These results indicate the increasing risk for a PTSD diag-
membership on PTSD symptom severity, F(3, 164) = nosis associated with increasing level of EMS. The PTSD
18.29, p \ .001, g2 = .25 (see Table 4). Least Significant positive rate of 33% in the Moderate-EMS group is compa-
Differences (LSD) post hoc tests examined group-level rable to the 39.1% lifetime prevalence of PTSD found among
differences in PTSD symptoms. The High-EMS subgroup all CSA-exposed women in the National Comorbidity Study
exhibited the greatest mean levels of PTSD symptom (Molnar et al. 2001). Clearly, however, the High- and Low-
severity (M = 52.0, SD = 14.8) and was significantly EMS groups comprise subgroups of CSA-exposed women
greater than each of the other three groups, ps \ .05. The who are at greater and lesser risk, respectively, than the
Moderate-EMS subgroup exhibited a moderate level of average woman.

123
570 Cogn Ther Res (2012) 36:560–575

Cluster Differences: Depressive Symptoms groups (PTSD-positive vs. PTSD-negative), Wilks’

and Trauma-Related Difficulties lambda = .58, v2 (15, N = 122) = 60.24, p \ .001, and
accounting for approximately 42% of the between groups
We next compared the cluster groups on two other related variability. Examination of the within groups structure
measures of psychopathology. Group differences on mea- coefficients revealed three predictors possessing large
sures of depressive symptoms (i.e., the BDI-II) and trauma- correlations with the discriminant function, namely Vul-
related behavioral and affective difficulties (i.e., the nerability to Harm (.67), Mistrust/Abuse (.63), and Emo-
Trauma Symptom Inventory) followed a similar pattern of tional Deprivation (.62). A second set of predictors were
group differences when compared to PTSD symptom lev- moderately correlated with the discriminant function.
els. Specifically, the Low-EMS CSA cluster did not differ These included Subjugation (.59), Social Isolation (.54),
from the non-CSA comparison group on these additional Defectiveness (.52), Emotional Inhibition (.49), and
measures of psychopathology. The High-EMS CSA cluster Abandonment (.47). Thus, women who met criteria for a
exhibited the greatest levels of depressive symptoms and presumptive PTSD diagnosis as defined in the current study
trauma-related symptomatology, and the Moderate-EMS were discriminated primarily by EMSs reflecting a per-
CSA cluster exhibited intermediate levels of these symp- ception of vulnerability to harm, interpersonal mistrust, and
toms (see Table 4). Given the high comorbidity between a belief that one’s emotional needs will not be met by
PTSD and depression, we also sought to determine whether others. Schemas related to a passive or deferential inter-
EMS-based group membership remained a significant personal style, a view of the self as defective, inadequate,
predictor of PTSD symptoms when controlling for non- or inherently different from others, and a tendency to
PTSD psychopathology. We examined this question by inhibit emotional expression appeared to contribute sec-
conducting a follow-up analysis in which PTSD symptoms ondarily to the differentiation of women who met pre-
were examined across EMS-based groups while BDI-II sumptive diagnostic criteria for PTSD.1
total symptoms were included as a covariate. Results of the
ANCOVA indicated that group membership was still sig-
nificantly associated with differences in PTSD symptom Discussion
severity level when accounting statistically for depressive
symptoms, F(3, 159) = 4.83, p = .003. Examination of Broadly stated, the aim of the present study was to examine
group means revealed the same pattern of PTSD symptom the relation between early maladaptive schemas and PTSD
level across groups, with the High-EMS group demon- symptoms among a sample of women with childhood
strating the greatest level of PTSD symptoms, the Moder- sexual abuse histories. The primary finding was that
ate-EMS group displaying a moderate level, and the women who have experienced childhood sexual abuse
Low-EMS group demonstrating the lowest level of PTSD differ, in some cases drastically, in regard to the degree to
symptoms. which they endorse beliefs reflective of early maladaptive
schemas. Specifically, our findings indicate that women
Discrimination of Presumptive PTSD Diagnostic Status with CSA histories comprise a heterogeneous group con-
with Early Maladaptive Schemas sisting of subgroups that can be distinguished meaningfully
by the severity of cognitive schema endorsement. Impor-
Finally, we wished to examine which schemas would be tantly, our study enabled identification of a significant
most efficient in discriminating women who met pre- minority of CSA survivors with elevated schema
sumptive diagnostic criteria for PTSD from those who did endorsement across content domains who also experience
not. Specifically, it was expected that the schemas of high levels of PTSD symptoms. This important subgroup
Vulnerability to Harm and Mistrust/Abuse would emerge as
significant predictors of PTSD presumptive diagnostic 1
As an adjunctive analysis, we conducted a single-entry linear
status. To examine this question, descriptive discriminant multiple regression in which all 15 schemas were entered as
analysis, a variable-centered analytic strategy, was con- predictors of the continuous total PTSD symptom scores. Results
ducted to predict presumptive PTSD diagnostic status suggested that the overall model significantly predicted PTSD
symptom severity, F(15, 106) = 4.21, p \ .001, adjusted R2 = .29,
(PTSD-positive vs. PTSD-negative) among the sample of
and indicated that Vulnerability to Harm (b = .31, p = .003) and
CSA survivors. Groups were defined based on the previ- Mistrust/Abuse (b = .27, p = .03) emerged as the only unique
ously described procedure used to assign a presumptive predictors of PTSD symptom severity when controlling for the
PTSD diagnosis on the basis of participants’ responses on collinearity among the schema subscales. Thus, results are largely
consistent with the discriminant analysis findings which revealed that
the PPTSD-R. Predictor variables were each of the 15 EMS
these two schema subscales were most strongly correlated with the
subscales. The discriminant function analysis revealed a discriminant function differentiating women with and without a
significant multivariate dimension discriminating the two presumptive PTSD diagnosis.

123
Cogn Ther Res (2012) 36:560–575 571

would be obscured by methods using a more traditional, prediction and with findings of Lumley and Harkness
variable-oriented approach, and as such, the current find- (2007), schemas relating to themes of Vulnerability to
ings emphasize the importance and utility of using a per- Harm and Mistrust/Abuse emerged as important predictors
son-centered approach to the examination of EMS among of presumptive PTSD diagnostic status. These schemas
CSA survivors. refer to beliefs that others may be expected to intentionally
The finding that differences among CSA survivors’ hurt, abuse, or humiliate, and an exaggerated belief that
schema profiles were largely in the severity of schema catastrophe or harm can unpredictably strike at any time
scores, rather than in the overall shape or distribution of the (Young and Brown 1994). As previously described, these
EMS profile, was, however, unexpected. Based on research schemas have demonstrated associations with anxious
demonstrating that both CSA experiences and survivors’ symptomatology in a prior study (Lumley and Harkness
responses are quite heterogeneous (Briere and Jordon 2007), and are theoretically similar to contemporary cog-
2004), we had expected variability in schema profiles as nitive models of PTSD that explicitly focus on cognitive
well. Instead, the profiles of all three CSA survivor sub- themes such as safety and trust (Resick and Schnicke
groups were similar in shape and distribution, and the three 1993). Emotional Deprivation, which reflects a belief that
CSA subgroups’ profiles were similar in shape to that of one’s emotional needs will not be met by others, also
the non-CSA comparison group. Further, the schema pro- emerged as a strong predictor of PTSD status. This schema,
files for the Low and Moderate-EMS groups were largely although not predicted a priori, makes sense given that this
‘‘flat’’, i.e., showed minimal variability across schemas. schema is thought to develop in the presence of a devel-
The High-EMS group showed some variability, but even in opmental environment that is perceived as cold, removed,
this subgroup, the relative schema elevations were dis- or not providing adequate care and nurturance (Young and
tributed across three of the five schema domains. These Brown 1994). Given that social support is an important
results thus suggest that CSA survivors are distinguished protective factor against PTSD among child sexual abuse
primarily by a generalized elevation of their maladaptive survivors (Hyman et al. 2003), the observed relationship
schemas, rather than by unique schema profiles. between a perception that one is inherently lacking in
Consistent with cognitive models of PTSD (e.g., Ehlers needed emotional support from others to PTSD diagnostic
and Clark 2000; Foa and Rothbaum 1998; Resick and status makes intuitive sense.
Schnicke 1993), our findings provide strong evidence that The other subscales that emerged as moderate predictors
Young’s early maladaptive schemas are related to PTSD of PTSD status, Subjugation, Social Isolation, Defective-
symptomatology. Level of schema elevation, as displayed ness, Emotional Inhibition and Abandonment, collectively
by the three CSA survivor subgroups, was linearly related to reflected a passive or deferential interpersonal style, a view
level of PTSD symptoms and to presumptive PTSD diag- of the self as defective, inadequate, or inherently different
noses. That is, the CSA survivors who most strongly from others, and a tendency to inhibit emotional expression.
endorsed maladaptive beliefs across various schema These intra- and interpersonal themes make sense given
domains had the highest level of PTSD symptoms and were other literature on PTSD. For example, previous research
also more likely to meet criteria for PTSD than CSA sur- provides evidence that emotional inhibition (termed emo-
vivors who endorsed these maladaptive beliefs to a lesser tional nonacceptance) is associated with increased rates of
degree. Importantly, this group also had levels of symptoms experiential avoidance and subsequent posttraumatic stress
and diagnoses commensurate with those found in clinical symptomatology among individuals with maltreatment
samples (Lauterbach and Vrana 1996), which is notable histories (Gratz et al. 2007; Tull et al. 2007). Taken toge-
given that the current data were drawn from a non-clinical ther, these findings suggest that beliefs and cognitions
sample of college women. This finding further underscores relating to the inhibition of emotional expression and
the utility of the person-centered approach when examining excessive yielding to others’ desires, preferences, or needs
EMS and PTSD, as it allowed for identification of this are relevant areas when considering the PTSD symptoms of
particularly at-risk group of CSA survivors. CSA survivors, and may be relevant to the emergence and/
A complementary perspective on the relationship of or maintenance of PTSD symptomatology.
schemas to PTSD symptoms was provided by the use of a An interesting finding emerged when we compared CSA
variable-centered analytic approach, the descriptive dis- survivors who exhibited low levels of early maladaptive
criminant analysis. Whereas the cluster analysis findings schemas to women without CSA histories. With the
suggest that elevated schemas across the entire range of exception of emotional neglect, the CSA survivors in the
content domains increase risk for PTSD, the discriminant group characterized by the lowest EMS scores endorsed
analysis showed that across the three schema-based sub- greater levels of abuse relative to the non-CSA comparison
groups of CSA survivors, certain schemas were more group. This makes sense given that different forms of abuse
predictive of PTSD than others. Consistent with our tend to co-occur and that our sample was selected on the

123
572 Cogn Ther Res (2012) 36:560–575

basis of CSA history. However, it is interesting to note that criteria, for example, including in the cluster analysis
these two groups did not differ in regard to EMS or PTSD women who endorsed experiencing any form of childhood
symptom levels. This may suggest the presence of pro- maltreatment, including emotional or physical abuse. Dif-
tective factors or characteristics that mitigate the influence ferent inclusion criteria may have resulted in greater het-
of negative childhood experiences upon the later develop- erogeneity in our sample and resultantly, may influence the
ment of maladaptive cognitive schemas. Factors such as subgroup structure identified by the cluster analysis. It is
social support, positive early relationships, secure attach- also possible, perhaps even likely, for the relationship
ment to primary caregivers, or individual differences in between early maladaptive schemas and PTSD symptoms
temperament and personality may be plausible factors to be influenced by factors such as age and time since
affecting the relationship between CSA and other forms of trauma. Thus, it would be a relevant extension of the cur-
maltreatment and early maladaptive schema development. rent study to examine this relationship across time and
Alternately, the level of maltreatment experienced by these across demographic groups.
women may have been sufficiently low or infrequent and Furthermore, the CSA survivors with the greatest levels
resultantly was not associated with maladaptive schema of EMS also endorsed the most emotional abuse in child-
content. hood, suggesting that women who experience sexual abuse
Perhaps not surprisingly, the women in our study with and emotional abuse during childhood may be at greatest
the most severe levels of schema endorsement also dis- risk for the eventual development of harmful levels of early
played the greatest vulnerability to psychopathology other maladaptive schemas. These findings are consistent with
than PTSD. Not only were they were more likely to meet other researchers who have noted the relationship between
criteria for PTSD, but they also exhibited the highest levels childhood emotional and sexual abuse and the development
of depressive symptoms (on average, they demonstrated of negative cognitive styles (e.g., Gibb 2002). However,
levels of depressive symptoms in the ‘‘moderate’’ range), this calls into question the ‘‘important ingredient’’ in pre-
and endorsed greater levels of trauma-related dysfunctional dicting EMS levels—it is plausible that emotional abuse
behaviors and affective symptomatology. This finding, drives the relationship between group membership and
although consistent with previous research highlighting the EMS severity. Future research may wish to control for this
influence of negative cognitive styles on depression potential confound or examine EMS among women
severity (Gibb 2002; Rose et al. 1994), also calls into selected on the basis of emotional abuse history to sepa-
question the specificity of the relationship between YSQ rately examine this question. Additionally, the present
subscales and PTSD observed in the present study. Previ- study had only limited information about the nature and
ous researchers have noted that the YSQ may function characteristics of the CSA history. Factors such as perpe-
more as a global measure of general cognitive vulnerability trator age, relationship to the perpetrator, duration and
(e.g., McGinn et al. 2005), and as such, the severity level of severity of the abuse, age of abuse, and revictimization
EMS may have less specificity in regard to various psy- information may all contribute important information to
chopathological outcomes. In this regard, our findings are understanding the differences among this sample in regard
consistent with general cognitive vulnerability models. to EMS.
However, it is also important to note that follow-up anal- Given the cross-sectional nature of our data, inferences
yses suggested that the association between EMS-based about the temporal relationships among our variables
group membership and PTSD symptomatology held when cannot be drawn. For example, are high levels of EMS
controlling statistically for depression severity, which predictive of PTSD symptom development, or do high
would suggest that the relationship between schema group levels of PTSD symptoms contribute to the development,
membership and PTSD symptomatology is not accounted maintenance, or exacerbation of rigid and maladaptive
for by the relationship between schema severity and psy- levels of cognitive schemas? Studies utilizing prospective
chopathology (measured here as depressive symptoms) designs are necessary to identify the directionality of the
more generally. An important question for future research associations between CSA, EMS, and posttraumatic stress
to examine is whether the specific relationship between symptomatology demonstrated in the current study.
EMS and PTSD symptoms can be replicated in additional Finally, uncertainty regarding the validity of the YSQ to
samples. measure early maladaptive schemas must be noted given
It is important to note that our findings are qualified by that a priming design was not utilized in this study to
several important limitations. First, limitations to the gen- ‘‘activate’’ schemas (which are posited to remain dormant
eralizability of the findings should be noted. The current until activated by an external or internal trigger). There-
sample included a sample of young adult female CSA fore, our findings may more accurately reflect cognitive
survivors in a university setting. It is possible that our products (similar to other measures of cognitive vulnera-
results may have differed if we used broader inclusion bility) as opposed to measuring actual schema activation.

123
Cogn Ther Res (2012) 36:560–575 573

More broadly, the self-report methodology used may also Briere, J., & Jordon, C. E. (2004). Violence against women: Outcome
be susceptible to a number of reporting biases. For exam- complexity and implications for assessment and treatment.
Journal of Interpersonal Violence, 19, 1252–1276. doi:10.1177/
ple, some respondents may under-endorse schema items 0886260504269682.
due to desirability concerns. Mood or current affective state Bromet, E., Sonnega, A., & Kessler, R. C. (1998). Risk factors for
could also bias recall for schema-congruent information, DSM-III-R posttraumatic stress disorder: Findings from the
possibly influencing response patterns. Given the possi- National Comorbidity Survey. American Journal of Epidemiol-
ogy, 147, 353–361.
bility for systematic influences on responses when EMS are Cahill, S. P., & Foa, E. B. (2007). Psychological theories of PTSD. In M.
measured via self-report, future research using alternative J. Friedman, T. M. Keane, & P. A. Resick (Eds.), Handbook of PTSD:
methodologies that are less susceptible to recall and Science and practice (pp. 55–77). New York: Guildford Press.
affective biases would be beneficial. Castille, K., Prout, M., Marczyk, G., Shmidheiser, M., Yoder, S., &
Howlett, B. (2007). The early maladaptive schemas of self-
To our knowledge this is the first study to document mutilators: Implications for therapy. Journal of Cognitive Psycho-
associations among PTSD and EMS among a sample of therapy, 21(1), 58–71. doi:10.1891/088983907780493340.
CSA survivors, and as such, the present study contributes Dalgleish, T. (2004). Cognitive approaches to posttraumatic stress
importantly to our understanding of maladaptive and disorder: The evolution of multirepresentational theorizing.
Psychological Bulletin, 130(2), 228–260. doi:10.1037/0033-2909.
enduring cognitive schemas about the self, world, and 130.2.228.
others among child sexual abuse survivors. Additional Dingemans, A. E., Spinhoven, P., & van Furth, E. F. (2007).
research is needed to replicate and extend the current Predictors and mediators of treatment outcome in patients with
findings, particularly in clinical and community samples binge eating disorder. Behaviour Research and Therapy, 45(11),
2551–2562. doi:10.1016/j.brat.2007.06.003.
and among increasingly diverse samples. Nonetheless, Ehlers, A., & Clark, D. M. (2000). A cognitive model of persistent
findings support the potential utility of schema-focused posttraumatic stress disorder. Behaviour Research and Therapy,
interventions particularly in combination with existing 38, 319–345. doi:10.1016/S0005-7967(99)00123-0.
cognitive and exposure-based treatments in an effort to Fink, L. A., Bernstein, D., Handelsman, J., & Lovejoy, M. (1995).
Initial reliability and validity of the childhood trauma interview:
modify existing core beliefs and decrease subsequent A new multidimensional measure of childhood interpersonal
posttraumatic stress symptomatology. Future investigation trauma. The American Journal of Psychiatry, 152, 1329–1335.
of schema specificity as it relates to CSA survivors may Foa, E. B., & Rothbaum, B. O. (1998). Treating the trauma of rape:
also serve as a useful component of early interventions Cognitive-behavioral therapy for PTSD. New York: Guildford
Press.
aimed at preventing posttraumatic stress or decreasing the Fraley, C., & Raftery, A. E. (1998). How many clusters? Which
likelihood of enduring and chronic symptoms. clustering method? Answers via model-based cluster analysis. The
Computer Journal, 41, 578–588. doi:10.1093/comjnl/41.8.578.
Acknowledgments This research was supported by the Verizon/ Fraley, C., & Raftery, A. E. (2006). MCLUST Version 3 for R:
HopeLine Student Fellowship for Domestic Violence Research Normal Mixture Modeling and Model-based Clustering, Tech-
awarded to the first author. nical Report No. 504, Department of Statistics, University of
Washington (revised December 2009).
Gibb, B. E. (2002). Childhood maltreatment and negative cognitive
styles: A quantitative and qualitative review. Clinical Psychol-
ogy Review, 22, 223–246. doi:10.1016/S0272-7358(01)00088-5.
References Gratz, K. L., Bornovalova, M. A., Delany-Brumsey, A., Bettina, N.,
& Lejuez, C. W. (2007). A laboratory-based study of the
American Psychiatric Association. (1994). Diagnostic and statistical relationship between childhood abuse and experiential avoidance
manual of mental disorders (4th ed.). Washington, DC: Author. among inner-city substance users: The role of emotional
Banfield, J. D., & Raftery, A. E. (1993). Model-based Gaussian and nonacceptance. Behavior Therapy, 38, 256–268. doi:10.1016/
non-Gaussian clustering. Biometrics, 49, 803–821. doi:10.2307/ j.beth.2006.08.006.
2532201. Harris, A. E., & Curtin, L. (2002). Parental perceptions, early
Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Manual for beck maladaptive schemas, and depressive symptoms in young adults.
depression inventory-II. San Antonio, TX: Psychological Cognitive Therapy and Research, 26, 405–416. doi:10.1023/A:
Corporation. 1016085112981.
Bernstein, D. P., & Fink, L. (1998). Childhood trauma questionnaire: Hicks, B. M., Markon, K. E., Patrick, C. J., Krueger, R. F., &
A retrospective self report: Manual. San Antonio, TX: Psycho- Newman, J. P. (2004). Identifying psychopathy subtypes on the
logical Corporation. basis of personality structure. Psychological Assessment, 16(3),
Breslau, N., Kessler, R. C., Chilcoat, H. D., Schultz, L. R., Davis, G. 276–288. doi:10.1037/1040-3590.16.3.276.
C., & Andreski, P. (1998). Trauma and posttraumatic stress Horowitz, M. (2001). Stress response syndromes (4th ed.). Northvale,
disorder in the community: The 1996 Detroit area survey of NJ: Aronson.
trauma. Archives of General Psychiatry, 55, 626–632. doi: Hyman, S. M., Gold, S. N., & Cott, M. A. (2003). Forms of social
10.1001/archpsyc.55.7.626. support that moderate PTSD in childhood sexual abuse survi-
Brewin, C. R., & Holmes, E. A. (2003). Psychological theories of vors. Journal of Family Violence, 18, 295–300.
posttraumatic stress disorder. Clinical Psychology Review, 23, Janoff-Bulman, R. (1992). Shattering assumptions: Toward a new
339–376. doi:10.1016/S0272-7358(03)00033-3. psychology of trauma. New York: Free Press.
Briere, J. (1995). Trauma symptom inventory professional manual. Kessler, R. C., McGonagle, K. A., Zhao, S., Nelson, C. B., Hughes,
Lutz, FL: Psychological Assessment Resources, Inc. M., Eshleman, S., et al. (1994). Lifetime and 12-month

123
574 Cogn Ther Res (2012) 36:560–575

prevalence of psychiatric disorders in the United States: Results Resick, P. A., & Schnicke, M. K. (1993). Cognitive processing
from the National Comorbidity Survey. Archives of General therapy for rape victims: A treatment manual. Thousand Oaks,
Psychiatry, 51, 8–19. CA: Sage Publications.
Lane, A., Young, R. L., Baker, A. E., & Angley, M. (2010). Sensory Rose, D. T., Abramson, L. Y., Hodulik, C. J., Halberstadt, L., & Leff,
processing subtypes in Autism: Association with adaptive G. (1994). Heterogeneity of cognitive style among depressed
behavior. Journal of Autism and Developmental Disorders, 40, inpatients. Journal of Abnormal Psychology, 103, 419–429. doi:
112–122. doi:10.1007/s10803-009-0840-2. 10.1037/0021-843X.103.3.419.
Lauterbach, D., & Vrana, S. (1996). Three studies on the reliability Schmidt, N. B., & Joiner, T. E. (2004). Global maladaptive schemas,
and validity of a self-report measure of posttraumatic stress negative life events, and psychological distress. Journal of
disorder. Assessment, 3, 17–25. Psychopathology and Behavioral Assessment, 26, 65–72. doi:
Lee, Z., Salekin, R. T., & Iselin, A. R. (2010). Psychopathic traits in 10.1023/B:JOBA.0000007457.95008.d2.
youth: Is there evidence of primary and secondary subtypes? Schmidt, N. B., Joiner, T. E., Young, J. E., & Telch, M. J. (1995). The
Journal of Abnormal Child Psychology, 38, 381–393. doi: schema questionnaire: Investigation of psychometric properties
10.1007/s10802-009-9372-7. and the hierarchical structure of a measure of maladaptive
Lumley, M. N., & Harkness, K. L. (2007). Specificity in the relations schemas. Cognitive Therapy and Research, 19, 295–301. doi:
among childhood diversity, early maladaptive schemas, and 10.1007/BF02230402.
symptom profiles in adolescent depression. Cognitive Therapy Skeem, J. L., Johansson, P., Andershed, H., Kerr, M., & Eno Louden,
and Research, 31, 639–657. doi:10.1007/s10608-006-9100-3. J. (2007). Two subtypes of psychopathic violent offenders that
Mason, O., Platts, H., & Tyson, M. (2005). Early maladaptive parallel primary and secondary variants. Journal of Abnormal
schemas and adult attachment in a UK clinical population. Psychology, 116, 395–409. doi:10.1037/0021-843X.116.2.395.
Psychology and Psychotherapy: Theory, Research and Practice, Steer, R. A., & Clark, D. A. (1997). Psychometric characteristics of
78(4), 549–564. doi:10.1348/147608304X21374. the Beck Depression Inventory-II with college students. Mea-
McGinn, L. K., Cukor, D., & Sanderson, W. C. (2005). The surement and Evaluation in Counseling and Development, 30,
relationship between parenting style, cognitive style, and anxiety 128–136.
and depression: Does increased early adversity influence symp- Stein, M. B., & Walker, J. R. (2000). Gender differences in
tom severity through the mediating role of cognitive style? susceptibility to posttraumatic stress disorder. Behavior
Cognitive Therapy and Research, 29, 219–242. doi:10.1007/ Research & Therapy, 38(6), 619–628. doi:10.1016/S0005-7967
s10608-005-3166-1. (99)00098-4.
Messman-Moore, T. L., & Coates, A. A. (2007). The impact of Timmons-Mitchell, J., Chandler-Holtz, D., & Semple, W. E. (1996).
childhood psychological abuse on adult interpersonal conflict: Post-traumatic stress symptoms in mothers following children’s
The role of early maladaptive schema patterns of interpersonal reports of sexual abuse: An exploratory study. American Journal
behavior. Journal of Emotional Abuse, 7(2), 75–92. doi: of Orthopsychiatry, 66, 463–467. doi:10.1037/h0080196.
10.1300/J135v07n02_05. Tull, M. T., Barrett, H. M., McMillan, E. S., & Roemer, L. (2007). A
Molnar, B. E., Buka, S. L., & Kessler, R. C. (2001). Child sexual preliminary investigation of the relationship between emotion
abuse and subsequent psychopathology: Results from the regulation difficulties and posttraumatic stress symptoms.
National Comorbidity Survey. American Journal of Public Behavior Therapy, 38, 303–313. doi:10.1016/j.beth.2006.10.
Health, 91, 753–760. doi:10.2105/AJPH.91.5.753. 001.
Mun, E. Y., Windle, M., & Schainker, L. M. (2008). A model-based Ullman, S. E., & Filipas, H. H. (2005). Ethnicity and child sexual
cluster analysis approach to adolescent problem behaviors and abuse experiences of female college students. Journal of Child
young adult outcomes. Development and Psychopathology, 20, Sexual Abuse, 14, 67–89. doi:10.1300/J070v14n03_04.
291–318. doi:10.1017/S095457940800014X. Waller, G., Kennerley, H., & Ohanian, V. (2007). Schema-focused
Nurius, P. S., & Macy, R. J. (2008). Heterogeneity among violence- cognitive behavioral therapy with eating disorders. In L. P. Riso,
exposed women: Applying person-oriented research methods. P. L. du Toit, D. J. Stein, & J. E. Young (Eds.), Cognitive
Journal of Interpersonal Violence, 23, 389–415. doi:10.1177/ schemas and core beliefs in psychiatric disorders: A scientist-
0886260507312297. practitioner guide (pp. 139–175). New York, NY: American
O’Dougherty Wright, M., Crawford, E., & Del Castillo, D. (2009). Psychological Association. doi:10.1037/11561-007.
Childhood emotional maltreatment and later psychological Waller, G., Meyer, C., & Ohanian, V. (2001). Psychometric
distress among college students: The mediating role of mal- properties of the long and short version of the Young Schema
adaptive schemas. Child Abuse and Neglect, 33, 59–68. doi: Questionnaire: Core beliefs among bulimic women and com-
10.1016/j.chiabu.2008.12.007. parison women. Cognitive Therapy & Research, 25, 137–147.
Petrocelli, J. V., Glaser, B. A., Calhoun, G. B., & Campbell, L. F. doi:10.1023/A:1026487018110.
(2001). Personality and affect characteristics of outpatients with Waller, G., Ohanian, V., Meyer, C., & Osman, S. (2000). Cognitive
depression. Journal of Personality Assessment, 77(1), 162–175. content among bulimic women: The role of core beliefs.
doi:10.1207/S15327752JPA7701_11. International Journal of Eating Disorders, 28, 235–241. doi:
Pinto-Gouveia, J., Castilho, A., & Cunha, M. (2006). Early 10.1037/11561-007.
maladaptive schemas and social phobia. Cognitive Therapy Young, J. E. (1994). Cognitive therapy for personality disorders: A
and Research, 30, 571–584. doi:10.1007/s10608-006-9027-8. schema-focused approach. Sarasota, FL: Professional Resource
Price, J. P. (2007). Cognitive schemas, defense mechanisms and post- Press. Revised edition.
traumatic stress symptomatology. Psychology and Psychother- Young, J. E. (1999). Cognitive therapy for personality disorders: A
apy: Theory, Research and Practice, 80, 343–353. doi:10.1348/ schema-focused approach (3rd ed.). Sarasota, FL: Professional
147608306X144178. Resource Press.
Raftery, A. E. (1995). Bayesian model selection in social research. Young, J. E., & Brown, G. (1994). Young Schema Questionnaire-
Sociological Methodology, 25, 111–163. doi:10.2307/271063. short form. New York, NY: Cognitive Therapy Center.

123
Cogn Ther Res (2012) 36:560–575 575

Young, J. E., Klosko, J. S., & Weishar, M. E. (2003). Schema Zlotnick, C., Zatriski, A. L., Shea, M. T., Costello, E., Begin, A.,
therapy: A practitioner’s guide. New York: Guilford Press. Pearlstein, T., et al. (1996). The long-term sequelae of sexual
Zinzow, H. M., & Jackson, J. L. (2009). Attributions for different abuse: Support for a complex posttraumatic stress disorder.
types of traumatic events and post-traumatic stress among Journal of Traumatic Stress, 9, 195–205. doi:10.1002/jts.
women. Journal of Aggression, Maltreatment, and Trauma, 18, 2490090204.
499–515. doi:10.1080/10926770903051025.

123