Journal of Personality Disorders, 34, 2020, 480

© 2020 The Guilford Press

DEVELOPMENTAL PATHWAYS TO
BPD-RELATED FEATURES IN ADOLESCENCE:
INFANCY TO AGE 15
Laura E. Brumariu, PhD, Margaret Tresch Owen, PhD,
Nazly Dyer, PhD, and Karlen Lyons-Ruth, PhD

The self-damaging behaviors central to borderline personality disorder

(BPD) become prominent in adolescence. Current developmental theories
cite both early family processes and childhood dysregulation as contributors
to BPD, but longitudinal data from infancy are rare. Using the National
Institute of Child Health and Human Development (NICHD) Study of
Early Child Care and Youth Development database (SECCYD; N = 1,364),
we examined path models to evaluate parent and child contributors
from infancy/preschool, middle childhood, and adolescence to adolescent
BPD-related features. In addition, person-centered latent class analyses
(LCA) investigated whether adolescent BPD-related features were more
strongly predicted by particular patterns of maladaptive parenting.
Path modeling identified unique influences of maternal insensitivity and
maternal depression on BPD-related features, first, through social-emotional
dysregulation in middle childhood, and second, through continuity from
infancy in maternal insensitivity and depression. LCA results indicated
that early withdrawn parenting was particularly predictive of BPD-related
features in adolescence. Results suggest multiple points of intervention to
alter pathways toward adolescent borderline psychopathology.

Keywords: borderline personality disorder, risky self-damaging behavior,

adolescence, maternal sensitivity, maternal depression

Adolescence is a critical time of transition from dependence on the family

to more autonomous functioning in the community. This transition involves
expansion of activities into a variety of new domains, including intimacy/
sexuality and work or career-related studies (e.g., Kobak, Zajac, Abbott, Zisk,

From Derner School of Psychology, Adelphi University, Garden City, New York (L. E. B.); School of
Behavioral and Brain Sciences, The University of Texas at Dallas (M. T. O.); Institutional Data Analytics,
University of Houston-Downtown (N. D); and Department of Psychiatry, Harvard Medical School at
Cambridge Health Alliance, Cambridge, Massachusetts (K. L.-R.).
This research was supported in part by a Harvard Psychiatry Department Livingston Award to Dr. Brumariu
and by NIMH grant R01MH06030 to Dr. Lyons-Ruth.
This study was based on data from the NICHD Study of Early Child Care. The NICHD support for the
data collection is gratefully acknowledged. We are thankful to families participating in the NICHD SEC-
CYD study.
Address correspondence to Dr. Laura E. Brumariu, Adelphi University, Gordon F. Derner School of Psychol-
ogy, 158 Cambridge Ave., Garden City, NY 11530. E-mail: lbrumariu@adelphi.edu

G4797_480.indd 1 6/8/2020 3:27:31 PM

 2 BRUMARIU ET AL.

& Bounoua, 2017). Notably, a variety of risky self-damaging behaviors are

also initiated during adolescence (e.g., DuRant, Smith, Kreiter, & Krowchuk,
1999). These increase in frequency over the adolescent years (Brener & Collins,
1998) and tend to cluster together (Pena, Matthieu, Zayas, Masyn, & Caine,
2012). Furthermore, these behaviors significantly threaten adolescent health
and well-being, result in high usage of public health services at enormous
costs, and place long-term constraints on further growth and development
(e.g., Youngblade, Curry, Novak, Vogel, & Shenkman, 2006). Studies indi-
cate that risky behaviors such as unsafe sex, self-harm, and illicit substance
use are key risk factors for incident disability-adjusted life years (DALYS),
contributing from 2% to 4% to DALYS among 10- to 24-year-olds globally
(Gore et al., 2011). Suicidality is of particular concern because it is among the
primary reasons for use of costly emergency and inpatient services (Peterson,
Zhang, Santa Lucia, King, & Lewis, 1996), and suicide is the second leading
cause of death among 15- to 24year-olds in the United States (Centers for
Disease Control and Prevention, 2017). When risky self-damaging behaviors
are clustered together, the morbidity, mortality, and social costs are likely to
be much higher.
Clustering of these risky self-damaging behaviors also characterizes the
adult psychiatric diagnosis of borderline personality disorder (BPD) (Ameri-
can Psychiatric Association [APA], 2013). Importantly, as noted in the fifth
edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-
5) criteria for BPD, individuals who engage in multiple types of risky self-
damaging behavior also tend to display particular forms of emotional and
social dysregulation, including labile moods, inappropriate intense anger, and
conflicted and unstable relationships, as well as identity diffusion, feelings of
emptiness, and fears of abandonment (APA, 2013). Not surprisingly, adoles-
cents with elevated BPD features show serious impairments in functioning
and high rates of psychiatric service use, whether or not they meet full criteria
for BPD (Chanen & Kaess, 2012). Thus, BPD-related features in adolescence,
features that include both risky self-damaging behaviors and particular forms
of social-emotional dysregulation, are of particular interest, because together
they index a prominent form of psychiatric disorder in adulthood. In the
current study, the very rich NICHD SECCYD database (NICHD Early Child
Care Research Network, 2001) was leveraged to develop a composite index
of BPD-related features at age 15, including risky self-damaging behaviors,
suicidal ideation, and relevant forms of social and emotional dysregulation,
from the extensive set of validated measures of adaptive and maladaptive
functioning included in that study.
The developmental trajectories toward adolescent BPD-related features
have been a topic of considerable interest, as reviewed below (e.g., Michon-
ski, Sharp, Steinberg, & Zanarini, 2013; Sharp, Steinberg, Temple & Newlin,
2014; Stepp et al., 2014; Vaillancourt et al., 2014; Wolke, Schreier, Zanarini,
& Winsper, 2012). In particular, increased understanding of the developmental
precursors of adolescent BPD-related features in infancy, preschool, and middle
childhood would facilitate earlier identification and treatment of children at
risk, before the increase in morbidity and mortality evident in adolescence.

G4797_480.indd 2 6/8/2020 3:27:31 PM

 DEVELOPMENTAL PATHWAYS TO BPD FEATURES 3

Theories of the etiology of BPD include complex transactions between

biological vulnerabilities and family environment, including characteristics of
the child (e.g., impulsivity, negative affectivity) and the family environment
(e.g., ineffective parenting, disorganized attachment, invalidation of emo-
tions) (Crowell, Beauchaine, & Linehan, 2009; Gunderson & Lyons-Ruth,
2008). The influence of family environment has been theorized to begin early
in development by a number of authors (Carlson, Egeland, & Sroufe, 2009;
Cole, Llera, & Pemberton, 2009; Fonagy & Luyten, 2009; Gunderson &
Lyons-Ruth, 2008). In one of the more detailed theoretical accounts, Fonagy
and Luyten (2009) have speculated that the early caregiving context and dis-
ruptions in attachment relationships lead to poor self–other differentiation
that perpetuates a combination of low thresholds for activation of the attach-
ment system under stress, in concert with low thresholds for deactivation of
the capacity for controlled mentalization. These low thresholds infringe upon
one’s capacity to navigate and understand interpersonal relationships. In a
number of contemporary theories (see Winsper, 2018), the negative effects of
poor early parental regulation, combined with child temperamental vulner-
ability, impede the development of emotional regulation ability in particular
(Cole et al., 2009; Crowell et al., 2009; Linehan, 1993). By middle childhood,
difficulties in emotion regulation also lead to disturbed peer relationships.
These difficulties in middle childhood are viewed as further linking earlier
risk factors to the development of BPD features in adolescence. This general
developmental model provides one potential account to explain the difficult
interpersonal cycles, affective dysregulation, and high levels of impulsivity,
as well as the painful inner experiences of aloneness and identity diffusion
of individuals experiencing BPD. Thus, theoretically, BPD-related features
are viewed as rooted in family processes that affect early social-emotional
regulation, rather than emerging in adolescence primarily in response to new
developmental stressors. In the current work, videotaped assessments of par-
enting sensitivity from infancy to adolescence were available and included in
longitudinal modeling to assess the contributions of parenting over time to
BPD-related features at age 15.
While prior theoretical models have not focused on a potential contribu-
tion of maternal depression, the developmental psychopathology literature
points to maternal depression as a contributor to particular self-damaging
behaviors in adolescence, as well as to BPD in adulthood. Maternal depres-
sion is associated with an increased risk of suicidal ideation in offspring
(Hammerton, Zammit, Thapar, & Collishaw, 2016), as well as with disturbed
eating behaviors (Tafà et al., 2017). In addition, children from adverse family
backgrounds, including a history of maternal depression, are at increased risk
for behavioral and emotional dysregulation at age 11 (Winsper, Zanarini, &
Wolke, 2012). Finally, studies have shown significant effects of parental depres-
sion on offspring substance use (Campbell, Morgan-Lopez, Cox, & McLoyd,
2009). These effects of parental depression may be understood theoretically as
operating through both biological and social mechanisms (e.g., Crowell et al.,
2009). Family association studies have shown that first-degree relatives of
those with BPD are more likely to have affective disorders, as well as increased

G4797_480.indd 3 6/8/2020 3:27:31 PM

 4 BRUMARIU ET AL.

rates of BPD (Zanarini, Barison, Frankenburg, Reich, & Hudson, 2009). In

addition, parental depression has repeatedly been associated with impaired
parenting, including both less parental structure and support and increased
parental hostility (Goodman & Gotlib, 2002; Lyons-Ruth, Wolfe, Lyubchik,
Grogan, & Steingard, 2002). Thus, it is important to model how maternal
depression at particular developmental stages may contribute to pathways
to BPD-related features in adolescence. The SECCYD database also includes
repeated assessments of maternal depression over time, so that longitudinal
path models in the current study also assess its contributions.
Particular child behaviors have also been proposed as possible early pre-
cursors of BPD-related features. In the infancy period, disorganized attachment
has been shown to contribute to prediction of BPD features in early adulthood
(Carlson et al., 2009; Lyons-Ruth, Bureau, Holmes, Easterbrooks, & Brooks,
2013). In infancy, the term disorganized refers to the apparent lack of a consis-
tent way of organizing attachment responses to the parent when under stress
(Main & Solomon, 1990). Starting at age 3, controlling attachment patterns
also emerge in which the child actively attempts to control or direct the par-
ent’s attention and behavior in either a punitive or caregiving manner (Main &
Cassidy, 1988). Lyons-Ruth et al. (2013) found that disorganized-controlling
attachment patterns observed at 8 years of age added to prediction of BPD
features at age 19. These controlling patterns are also of interest because Kobak
et al. (2017) found that, among male adolescents, controlling-caregiving pat-
terns of interaction with parents at age 13 predicted increased levels of risky
self-damaging behavior by age 15. Finally, Lyons-Ruth, Brumariu, Bureau,
Hennighausen, and Holmes (2015) found that elevated BPD features, as well
as increased incidence of suicidality/self-injury specifically, were associated
with controlling-caregiving patterns of interaction with mothers at age 19.
Both disorganized attachment at age 15 months and controlling forms of
attachment at age 3 years were assessed in the SECCYD and are included in
the longitudinal path models assessed in the current study.
Major theoretical models (e.g., Crowell et al., 2009; Fonagy & Luyten,
2009; Gunderson & Lyons-Ruth, 2008) also suggest that peer interactions
indicating heightened risk for later BPD are likely to be identifiable by mid-
dle childhood. Regarding child behavioral precursors in middle childhood,
the Avon Longitudinal Study of Parents and Children (ALSPAC) found that
increased experiences of bully-victimization between 4 and 10 years of age
were precursors to affective dysregulation, impulsivity, and interpersonal dys-
function at age 11, assessed by interview (Wolke et al., 2012). These difficul-
ties are quite similar to the forms of behavioral and emotional dysregulation
observed as criteria for BPD in adulthood and are thought to be likely precur-
sors of BPD-related features in adolescence (Wolke et al., 2012). In addition,
other forms of childhood dysregulated behavior between 4 and 8 years of age
(i.e., negative emotionality, externalizing problems, and hyperactivity) also
predicted affective dysregulation, impulsivity, and interpersonal dysfunction
at age 11 (Winsper et al., 2012). In the McMaster Teen Study, childhood rela-
tional aggression by boys at ages 10 to 14, and both physical and relational
aggression by girls at age 14, were related to emotional dysregulation at age
14 (Vaillancourt et al., 2014).

G4797_480.indd 4 6/8/2020 3:27:31 PM

 DEVELOPMENTAL PATHWAYS TO BPD FEATURES 5

The ALSPAC study also found that being bullied and victimized in middle
childhood (7–10 years) was specifically associated with self-harm at 16 to 17
years (Lereya et al., 2013). Similarly, Klomek, et al. (2009) found that children
involved in bullying, both bully/victims and chronic victims, were at increased
risk for suicidality/self-injury in preadolescence. Finally, studies have confirmed
that peer victimization is often associated with substance use (Earnshaw et al.,
2017) and disordered eating (Lee & Vaillancourt, 2018). Thus, peer experi-
ences of bully-victimization in middle childhood contribute to emotional and
behavioral dysregulation. In the SECCYD study, several relevant indices of
social-emotional dysregulation in middle childhood were available, including
intense anger, suicidal ideation, impulsivity, peer conflict, and victimization.
Based on the above literature, a composite measure of middle childhood
social-emotional dysregulation was included in the longitudinal model as one
potential mediator of the effects of early insensitive parenting and disturbed
attachment on adolescent BPD-related features.
In addition to general models emphasizing maladaptive parenting,
increasing empirical work has looked more specifically at the particular forms
of parenting disturbances related to the emergence of BPD-related features in
adolescence. A variety of studies have linked aspects of parenting to adolescent
self-damaging behavior. Exposure to harsh parenting has emerged as one fac-
tor related to alcohol use and risky sex (Alati et al., 2014; Guilamo-Ramos
et al., 2012). Other studies have found a relation between lack of parental
support and self-damaging behaviors such as suicidal ideation, substance use,
and disturbed eating (Hammerton et al., 2016; Tafà et al., 2017).
Several relevant studies have also used direct observation in short- or
long-term longitudinal designs to assess the contributions of particular aspects
of early parenting and attachment quality to the development of BPD features
in adulthood. Carlson et al. (2009) reported a relation between observed harsh
parenting at age 3 and the extent of BPD features assessed at age 20. Lyons-
Ruth et al. (2013) found that observed maternal withdrawal in infancy was a
potent predictor of both suicidality/self-injury specifically and also of overall
BPD features at age 19. Analyzing expressed emotion in parental speech sam-
ples, Belsky et al. (2012) found that exposure to harsh treatment in the family
environment through age 10 predicted affective dysregulation, impulsivity, and
interpersonal dysfunction at age 12. A meta-analysis by Winsper et al. (2016)
showed an approximately threefold increase in risk of BPD-related features
in children and adolescents exposed to maternal hostility or verbal abuse.
Importantly, Stepp et al. (2014) demonstrated reciprocal associations
between parenting behavior and adolescent BPD symptoms, in that low paren-
tal warmth and harsh punishment predicted subsequent increases in adolescent
BPD features. In turn, adolescent BPD features predicted subsequent decreases
in parental warmth and increases in harsh punishment. In a narrative review of
the parenting behaviors of mothers who themselves had BPD, Stepp, Whalen,
Pilkonis, Hipwell, and Levine (2011) suggested that mothers with BPD may
oscillate between extreme forms of hostile control and passive aloofness in
their interactions with their children. Therefore, the existing literature sug-
gests that both withdrawing (lack of parental support) and harsh or hostile
parenting, either in oscillation or in different developmental periods, may

G4797_480.indd 5 6/8/2020 3:27:31 PM

 6 BRUMARIU ET AL.

make contributions to the emergence of BPD-related features in adolescence.

To address this question of whether particular aspects of insensitive parent-
ing may be differentially important in pathways to BPD, latent class analysis
was also used to identify distinct classes of parents exhibiting qualitatively
different profiles of early parenting. Those early parenting classes were then
assessed for how different patterns of parenting might relate to BPD-related
features in adolescence.

AIMS OF THE CURRENT STUDY

In the current study, we leverage the unusually comprehensive longitudinal

database from the NICHD SECCYD (N = 1,364). The SECCYD is one of the
only large-scale longitudinal studies to include repeated videotaped assess-
ments of mother–child interaction from infancy to age 15, as well as video-
taped assessments of early attachment behaviors. As noted, the database also
includes assessments of a number of other parental and child factors thought
to be theoretically central to the emergence of BPD-related features at age 15.
The SECCYD database also provides assessments of bully-victimization, peer
conflict, intense anger, impulsivity, and suicidality in middle childhood. Finally,
although no validated BPD scales for children and adolescents were available
at the time of the SECCYD, a range of risky or self-damaging behaviors and
their associated forms of social and emotional dysregulation were assessed at
age 15 using well-validated measures of substance abuse, risky sexual behavior,
disturbed eating, suicidal thoughts, peer conflict, bully-victimization, impulsiv-
ity, and uncontrolled, intense anger.
The first aim of the study was to evaluate a path model of longitudinal
contributors to BPD-related features at age 15. On the basis of the literature
reviewed above, we expected that early maternal insensitivity and depression
in the infancy and preschool periods would lead to difficulties in emotion
regulation and conflicted peer relations in middle childhood. In turn, such
difficulties in middle childhood would add to the likelihood of the child exhib-
iting BPD-related features in adolescence. In addition, if maternal depression
or insensitivity were not limited to early childhood but continued into middle
childhood and adolescence, such continued parental dysfunction would be
likely to contribute further to BPD-related features at age 15.
Thus, the following hypotheses were evaluated through path modeling:
(a) Insensitive parenting, maternal depression, and child disorganized attach-
ment behavior in the first 5 years of life will contribute to BPD-related features
in adolescence; (b) the influence of early parenting insensitivity and depres-
sion will be mediated in part through the continuation of these maladaptive
processes through middle childhood and adolescence; and (c) the influence
of early parenting insensitivity and depression will also be mediated in part
through their influence on the emergence of social-emotional dysregulation
in middle childhood.
The second aim of the study was to supplement path models based
on the continuous index for maternal sensitivity with person-centered latent
class models. The central parenting variable in the SECCYD was a composite

G4797_480.indd 6 6/8/2020 3:27:31 PM

 DEVELOPMENTAL PATHWAYS TO BPD FEATURES 7

measure of maternal sensitivity-insensitivity, with sensitivity defined as “being

aware of and correctly interpreting the child’s signals, and responding appro-
priately and promptly” (Ainsworth, Bell & Stayton, 1974, p. 129). How-
ever, maternal insensitivity can be manifested in varied ways, including harsh
parenting and withdrawn parenting, that may make different contributions
to BPD-related features. Person-centered models are better able to identify
distinct groups of parents who differ in their profiles of parental behavior.
Evidence suggests that both hostile/harsh and withdrawn parenting profiles
are relevant for the development of BPD-related features (Lyons-Ruth et al.,
2013; Stepp, Lazarus, & Byrd, 2016). By age 54 months, the SECCYD parental
interaction assessments included six scales, only three of which were included
in the single composite score for sensitivity (see Methods below). Hypotheses
for the LCA were the following: (a) There would be several distinct patterns
of early parenting in the sample that varied in hostility and withdrawal, and
(b) both hostile and withdrawn patterns would predict later social-emotional
dysregulation in middle childhood and BPD-related features at age 15.

METHOD
PARTICIPANTS
Participants were enrolled in the NICHD Study of Early Child Care and Youth
Development (SECCYD). Children and their families were recruited from 10
different sites across the United States when the target child was an infant (see
NICHD Early Child Care Research Network, 2001, for recruitment details).
For the present study, data were utilized from all phases of data collection:
Phase 1 (birth to age 3 years, N =1,364), Phase 2 (54 months through first
grade, N = 1,226), Phase 3 (second through sixth grades, N = 1,061), and
Phase 4 (age 15 years, N = 1,009). Mothers provided demographic data.
Of the recruited sample, 48.3% of target children were female and 19.6%
were ethnic minority children (White, n = 1,097, African American, n = 176,
and other ethnicity, n = 91). Mothers’ levels of education ranged from 7 to
21 years M (SD) = 14.23 (2.51). Approximately 10.3% (n = 139) had not
completed high school, 21% (n = 287) had completed high school or obtained
a GED, 33.4% (n = 455) had completed some college, and 35.3% (n = 482)
had a college or graduate degree. Presence or absence of the mother’s partner
in the home was examined at ages 1 month, 36 months, and 15 years, and the
variable “single parenting” indexed the number of those three assessments in
which the partner was living in the home. Approximately 8.7% (n = 118) did
not have a partner in the home at any of these time points; 22.2% (n = 303)
had a partner at one time point, 24.2% (n = 330) had a partner at two time
points, and 44.9% of mothers (n = 613) had a partner at all three time points.
Family income-to-needs ratio was assessed at ages 1 month, 36 months, 54
months, Grade 3, Grade 5, Grade 6, and 15 years. Income-to-needs ratios
were highly correlated (rs range from .49 to .89, p < .001); therefore, a mean
score across age was used in analyses (M = 3.80).
At age 15, measures of adolescent outcomes were obtained for 984 youth
(72% of the original recruitment sample). Compared with adolescents who

G4797_480.indd 7 6/8/2020 3:27:31 PM

 8 BRUMARIU ET AL.

did not participate at age 15, adolescents who participated were more likely
to have families with higher income-to-needs ratios, t(833.05) = 7.30, p < .001
(Ms = 2.87 and 4.15, respectively). Compared with mothers who were no lon-
ger participants at age 15, mothers who did participate had more education,
t(1361) = 5.07, p < .001 (Ms = 13.68 and 14.45, respectively). There were
no differences in retention by children’s ethnicity, White versus not-white:
χ2(1) = 1.72, p =.19. Girls were more likely to have data at age 15 than boys,
χ2(1) = 5.05 p < .05, 75% of recruitment sample versus 69.5%, respectively.

MEASURES
BPD-Related Features at Age 15 (BPDRF-15). A composite measure of BPD-
related features at age 15 was created using well-validated measures assessing
substance abuse, risky sexual behavior, disturbed eating, suicidal thoughts,
disturbed peer relationships, impulsivity, and uncontrolled, intense anger (see
Table 1 for descriptive statistics for individual indicators). To the extent pos-
sible, indicators were chosen to represent the BPD-related criteria represented
in the adult DSM-5 criteria in order to maximize the relevance of findings
to the understanding of BPD. Regarding indicators of risky behaviors, sub-
stance abuse was scored from the teen’s reported use of alcohol, marijuana,
and tobacco on the Risky Behavior Questionnaire (adapted from Conger &
Rand, 1994). Disordered eating was captured using the Eating Attitudes Test
(Garner, Olmsted, Bohr, & Garfinkel, 1982), indexing bingeing and restrict-
ing behavior. Risky sexual behavior was scored using teen responses on the
Risky Behavior Questionnaire. Self-reported impulsivity was measured with
scores on the eight-item Weinberger Adjustment Inventory (Weinberger &
Schwartz, 1990; alpha = .82; sample item: “I’m the kind of person who will
try anything once, even if it’s not that safe.”).
Suicidal thoughts or behavior at age 15 were measured from the three
suicidal thoughts/behavior items in the Youth Self Report (YSR; Achenbach,
1991b). In middle childhood (see below), these items were from the Child
Behavior Checklist (CBCL; Achenbach, 1991a), administered to the study
children’s mothers in Grades 3, 4, 5, and 6. Across these multiple assessments,
there were 105 children with at least one report of suicidal thoughts and/or
behavior.
Lack of control of anger or inappropriate intense anger was measured
from three relevant items on the YSR at age 15: “gets in many fights,” “physi-
cally attacks people,” and “temper tantrums/hot temper.” In middle childhood,
lack of control of anger or inappropriate intense anger was measured from
the same items on the CBCL. One hundred eighty-five children had reports
of many fights and physical attacks on others.
Unstable and conflicted relationships were assessed based on repeated
child reports of victimization on the Peer Social Support, Bullying, and Vic-
timization questionnaire (Kochenderfer & Ladd, 1996), available at age 15
and in Grades 3, 5, and 6, and on scores on the conflict and betrayal subscale
of the Friendship Quality Questionnaire (Clark & Ladd, 2000), administered
to children at age 15 as well as in Grades 3, 4, 5, and 6. An overall compos-
ite measure for BPD-related features at age 15 (BPDRF-15) was created by

G4797_480.indd 8 6/8/2020 3:27:31 PM

 DEVELOPMENTAL PATHWAYS TO BPD FEATURES 9

TABLE 1. Descriptive Statistics for Individual Unstandardized Indicators of Social-

Emotional Dysregulation in Middle Childhood and BPD-Related Features at Age 15
Variable n M (SD) Minimum Maximum
Grade 3
Suicidality 1026 .02 (.12) 0 2
Anger 1026 .18 (.27) 0 2
Conflict 1010 1.52 (.65) 1 5
Victimization 994 1.84 (.79) 1 5
Grade 4
Suicidality 1022 .01 (.11) 0 2
Anger 1022 .15 (.26) 0 2
Conflict 1018 1.52 (.66) 1 5
Grade 5
Suicidality 1020 .02 (.10) 0 1.5
Anger 1020 .14 (.25) 0 1.67
Conflict 993 1.47 (.59) 1 4.25
Victimization 987 1.80 (.77) 1 5
Grade 6
Suicidality 1023 .02 (.11) 0 1
Anger 1023 .13 (.24) 0 1.33
Conflict 999 1.43 (.43) 1 5
Victimization 990 1.76 (.72) 1 5
Age 15
Suicidality 956 .04 (.16) 0 1.5
Anger 957 .04 (.12) 0 .67
Conflict 973 1.44 (.58) 1 4.25
Victimization 956 .49 (.63) 0 4
Impulsivity 957 3.51 (.90) 1 5
Sexual risk taking 951 .22 (.56) 0 2
Disordered eating 957 5.49 (6.43) 0 59
Tobacco use 955 .17 (.49) 0 2
Alcohol use 954 .34 (.64) 0 2
Marijuana use 955 .14 (.45) 0 2

standardizing the scores of the 10 indicators above and then averaging the z
scores of the 10 indicators (α = .70).

Social-Emotional Dysregulation in Middle Childhood (SEDR-MC). In middle

childhood, many forms of risky self-damaging behavior are not yet appar-
ent (e.g., risky sex, risky driving, substance use, disordered eating). Thus, the
middle childhood composites were composed primarily of aspects of social-
emotional dysregulation (see Table 1 for descriptive statistics for individual
indicators). For Grades 3, 5, and 6, four relevant indicators of social-emotional
dysregulation were available, including anger, suicidality, conflict, and victim-

G4797_480.indd 9 6/8/2020 3:27:32 PM

 10 BRUMARIU ET AL.

ization, as described above. At Grade 4, anger, suicidality, and conflict were

available, but not victimization. For Grade 1, only anger and suicidality were
available, so Grade 1 was not included in the analyses. For middle child-
hood, the three or four available indicators were composited at each age by
standardizing the scores and then averaging the z scores of the indicators to
form a composite score for social emotional dysregulation. Social emotional
dysregulation scores for each grade showed good stability from Grades 3 to
6, with rs ranging from .41 to .55, all ps < .001. The four composite scores
for Grades 3, 4, 5, and 6 were then averaged into a single composite score to
reflect overall middle childhood SEDR-MC (α = .79).

Maternal Sensitivity. Maternal sensitivity was indexed using composites of

observational ratings based on videotaped mother-child interaction tasks at 6,
15, 24, 36, and 54 months; at Grades 1, 3, and 5; and at 15 years. At ages 6,
15, and 24 months, the sensitivity composite was the sum of 4-point ratings
of three items: sensitivity to nondistress, positive regard, and intrusiveness
(reverse scored). At age 54 months and Grades 1, 3, and 5, the sensitivity
composite was the sum of 7-point ratings of supportive presence, respect
for autonomy, and hostility (reverse scored). At age 15, the sensitivity vari-
able was a composite of six 7-point ratings for validation/agreement, engage-
ment, inhibiting relatedness (reversed), hostility/devaluing (reversed), respect
for autonomy, and valuing/warmth. Interrater reliability for each rating item
above ranged from .68 to .89 (ns 195 to 259). The composite variables had
reliabilities ranging from .83 to .91 (αs from .79 to .88).
Maternal sensitivity variables across infancy/early childhood were sig-
nificantly associated, rs = .30 to .52, ps < .001, as were maternal sensitivity
variables across middle childhood, rs = .37 to .42, p < .001. Because we were
interested in maternal sensitivity at three key developmental periods (infancy/
preschool, middle childhood, and adolescence), maternal sensitivity composite
variables were averaged over each developmental period, α = .75 for infancy/
preschool years, α = .65 for middle childhood. Sensitivity at age 15 was a
single score.

Maternal Interaction at 54 Months. In addition to the widely used SECCYD

sensitivity composites described above, the six mother–child interaction ratings
assessed at 54 months were used for person-centered LCA to identify patterns
of early parenting. The rating items at 54 months included the three items
used to generate the sensitivity composites above, as well as three additional
items assessing cognitive stimulation, quality of assistance, and confidence.
All six items were used to more specifically characterize maternal behavior for
the LCA. Interrater reliability (n = 242) ranged from .78 to .87 for the three
additional items not reported above.

Maternal Depression. Mothers completed the Center for Epidemiologic Studies

Depression Scale (CES-D, Radloff, 1977) at 1 month, 6 months, 15 months,
24 months, 36 months, 54 months, first grade, third grade, fifth grade, sixth
grade, and age 15 years (αs = .88 to .92). The CES-D is a widely used 20-
item self-report scale used to measure current levels of depressive features,

G4797_480.indd 10 6/8/2020 3:27:32 PM

 DEVELOPMENTAL PATHWAYS TO BPD FEATURES 11

with well-established reliability and validity (Radloff, 1977).Items are rated

on a 4-point scale according to how often they were experienced over the
past week. Correlations of scores across time points ranged from .35 to .58,
ps < .001. Similar to maternal sensitivity, composite scores were created for the
infancy/early childhood period (1 to 54 months, rs = .39 to .58, all ps < .001,
α = .84) and the middle childhood period (first to sixth grades, rs =.45 to .58,
all ps < .001, α = .80). Age 15 was indexed by a single score.

Attachment Insecurity and Disorganization. Attachment was assessed at 15

months with the Strange Situation Procedure (SSP; Ainsworth, Blehar, Waters,
& Wall, 1978) and at 36 months with the modified Strange Situation Proce-
dure (SSPm; Cassidy, Marvin, & MacArthur Working Group, 1992). During
the SSP, the mother leaves and rejoins the infant twice, first leaving the infant
with a female stranger, then leaving the infant alone to be rejoined by the
stranger. The procedure is designed to be mildly stressful in order to increase
the activation of the infant’s attachment behavior. Videotapes were coded
for secure, avoidant, ambivalent, disorganized, and unclassifiable attachment
behavior (Ainsworth et al., 1978; Main & Solomon, 1990). Intercoder agree-
ment among three trained coders was 83% (Κ = .69; N = 1201), with 59.6%
secure, 13.4% avoidant, 6% ambivalent, and 18.4% disorganized. The 3.5%
in the unclassifiable group were combined with the disorganized category.
The SSPm at 36 months also consists of two separations and reunions,
with the second separation longer than in infancy (5 min) to be more stressful
for older children (Cassidy et al., 1992). For the four-category ABCD sys-
tem, intercoder agreement among three trained coders was 75.7% (Κ = .58;
N = 867), with 61.5% secure, 4.8% avoidant, 17.3% ambivalent, and 16.4%
disorganized/controlling. A number of studies have provided evidence for
adequate psychometric properties of the SSP and the SSPm (van IJzendoorn,
Schuengel, & Bakermans-Kranenburg, 1999), to capture overall quality of
attachment at each age, we used a three-level scale (1 = secure, 2 = organized
avoidant and ambivalent, and 3 = disorganized or controlling).

PLAN OF ANALYSIS
Missing data ranged from 0% to 34.6%. For control analyses and descriptive
statistics, only participants with data on the relevant variables were included,
and pairwise tests were used to retain maximum power (see ns in Tables 1
and 2). To retain power for the main analyses, both bivariate correlations
involving SEDR-MC and BPDRF-15 and subsequent path modeling of the
direct and indirect paths from predictor variables to SEDR-MC and BPDRF-
15 were conducted using full information maximum likelihood (FIML) for
analysis and estimation of missing data (AMOS software). FIML provides a
less biased estimate and is more efficient in handling missing data than are
more conventional methods (Schafer & Graham, 2002).
The sequence of analyses was as follows: First, we evaluated whether
BPDRF-15 was predicted by SEDR-MC. Second, we evaluated whether the
composite scores for maternal sensitivity and maternal depression across
infancy/early childhood, middle childhood, and at age 15 were related to

G4797_480.indd 11 6/8/2020 3:27:32 PM

 12 BRUMARIU ET AL.

SEDR-MC and BPDRF-15. We also examined associations between mother–

child attachment at ages 15 and 36 months and SEDR-MC and BPDRF-15.
Third, we conducted path models to evaluate the direct and indirect pathways
through which maternal sensitivity, maternal depression, and mother–child
attachment predicted the emergence of SEDR-MC and BPDRF-15. Finally,
to assess which profiles of early parenting disturbance were most predictive
of SEDR-MC and BPDRF-15, the 54-months ratings were used for person-
centered LCA, and latent classes were assessed by ANOVA for prediction of
SEDR-MC and BPDRF-15.

RESULTS
CONTROL VARIABLES AND DESCRIPTIVE STATISTICS
Preliminary analyses investigated whether demographic variables were related
to BPDRF-15. Because family income-to-needs was highly related to maternal
education (r = .55, p < .001), a single socioeconomic status (SES) score was
created by averaging the standardized scores of the two variables. Both higher
SES and single parenting were negatively correlated with BPDRF-15 (SES
r = −.18, p < .001; single parenting r = −.20, p < .001). Gender was unrelated
to BPDRF-15, t(982) = −0.35, p > .05. Ethnicity was related to BPDRF-15,
F(2, 981) = 3.15, p < .05, but the Bonferroni post hoc analysis showed no
significant differences among groups. Therefore, only SES and single parenting
were included as covariates in the path analyses.
Descriptive statistics for key variables are shown in Table 2. Among the
early childhood predictors, quality of attachment (security/insecurity/disorga-
nization) at 15 and 36 months was modestly related (r = .07, p < .05). Qual-
ity of attachment at 15 months was also modestly associated with maternal
sensitivity in early childhood (r = .08, p < .01), but not with early maternal
depression. Quality of attachment at 36 months was significantly associated

TABLE 2. Descriptive Statistics for Dependent and Independent Variables

Variable n M (SD) Minimum Maximum
1. Infancy/preschool maternal sensitivity 1306 12 (1.93) 4 17.50
2. Middle childhood maternal sensitivity 1115 17.05 (2.10) 3 21
3. Age 15 maternal sensitivity 898 31.13(5.05) 9 42
4. Infancy/preschool maternal depression 1363 9.89 (6.77) 0 40
5. Middle childhood maternal depression 1125 8.75 (7.04) 0 48
6. Age 15 maternal depression 973 10.48 (9.83) 0 54
7. Insecurity/disorganization (15 months) 1191 1.59 (.78) 1 3a
8. Insecurity/disorganization (36 months) 1140 1.55 (.76) 1 3a
9. SEDR-MC 1096 .0012 (.49) −.93 4.82b
10. BPDRF-15 984 −.0008 (.54) −.76 2.70b
Note. SEDR-MC = social-emotional dysregulation in middle childhood; BPDRF-15 = BPD-related features at age 15.
aCoded 1 = Secure; 2 = Organized insecure; 3 = Disorganized (Disorganized/controlling at 36 months); bComposite

scores for SEDR-MC and BPDRF-15 could be negative due to standardization.

G4797_480.indd 12 6/8/2020 3:27:32 PM

 DEVELOPMENTAL PATHWAYS TO BPD FEATURES 13

with both early sensitivity (r = .22, p < .001) and early maternal depression
(r = −.11, p < .001). Early maternal depression was negatively related to early
sensitivity (r = −.31, p < .001). (Associations at other ages are presented in the
context of path analyses below.)

BIVARIATE RELATIONS OF PREDICTORS WITH OUTCOMES

Early Attachment and Later Outcomes. Quality of attachment at 15 months
was significantly but weakly associated with BPDRF-15 (r = .07, p < .05) but
not with SEDR-MC (r = .01, p = ns). Quality of attachment at 36 months was
significantly associated with SEDR-MC (r = .10, p < .05) but not BPDRF-15
(r = .03, p = ns). Follow-up orthogonal comparisons indicated that these results
were driven by attachment disorganization rather than by attachment insecu-
rity: secure versus organized insecure, all ns; organized versus disorganized,
36 months to SEDR-MC, F(1, 1006) = 5.49, p < .05, η2 =.005; 15 months to
BPDRF-15, F(1, 931) = 3.59, p =.06, η2 = .004.

Maternal Sensitivity and Depression and Later Outcomes. Maternal sensitivity

at all three developmental periods was significantly related to both SEDR-MC
and BPDRF-15 (SEDR-MC: Infancy/preschool r = −.25, Middle Childhood
r = −.27, Age 15 r = −.17, all p < .001; BPDRF-15: Infancy/preschool r = −.19,
Middle Childhood r = −.17, Age 15 r = −.16, all p < .001). The same was
true of maternal depression (SEDR-MC: Infancy/preschool r = .24, Middle
Childhood r = .33, Age 15 r = .22, all p < .001; BPDRF-15: Infancy/preschool
r = .13, Middle Childhood r = .17, Age 15 r = .19, all p < .001).

SEDR in Middle Childhood and BPDRF-15. Finally, bivariate correlations in-

dicated that higher levels of SEDR-MC (Grades 3, 4, 5, and 6) were associated
with higher levels of BPDRF at age 15 (r = .26, p < .001). Thus, early forms
of social-emotional dysregulation in middle childhood predicted increased
BPD-related features in mid-adolescence.

PATH MODELING OF TRAJECTORIES FROM

EARLY CHILDHOOD TO BPDRF-15
The theoretical path model of predictors of BPDRF-15 is shown in Figure 1,
as represented by both solid and dotted lines. A path analysis assessed the
goodness of fit of the proposed model, including indirect effects from early
maternal sensitivity, early maternal depression, and early attachment quality
through middle childhood SEDR to BPDRF-15. Because attachment at 15
months was directly related only to BPDRF-15, we included only that direct
pathway. Because single parenting and SES were also significantly related to
BPDRF-15, we included those variables as covariates. Covariances among all
predictor variables were included in the model.
The model showed good fit to the data, χ2(36) = 194.97, p < .001;
CFI = .95, IFI = .95; TLI = .89, RMSEA = .06. Not surprisingly, with all
variables included in the model, several of the paths significant in bivariate
correlations did not make independent contributions to prediction, including

G4797_480.indd 13 6/8/2020 3:27:32 PM

 14 BRUMARIU ET AL.

FIGURE 1. Theoretical and final models indicating Pathways to RSDB at age 15.
Dotted lines represent paths in the theoretical model that were nonsignificant
in the final model; solid lines indicate significant paths in the final model.
SEDR-MC = social-emotional dysregulation in middle childhood;
BPDRF-15 = BPD-related features at age 15. ***p < .001. **p < .01. *p < .05.

quality of attachment at 15 and 36 months, maternal depression in infancy/

preschool, and family SES, as indicated by the dotted lines in Figure 1. Thus,
we trimmed the nonsignificant paths from the model (significant standard-
ized path estimates shown in solid lines in Figure 1). The final model showed
very good fit to the data, χ2(22) = 107.41, p < .001; CFI = .97, IFI = .97,
TLI = .93, RMSEA = .05, and was a better fit than the nontrimmed model
(delta CFI = .02), based on adequate cutoff differences proposed by Meade,
Johnson, and Braddy (2008).
Because the distributions of some variables in the model deviated from
normality (maternal sensitivity in middle childhood, skewness  =  −1.35;
kurtosis = 3.83; social-emotional dysregulation in middle childhood, skew-
ness = −2.44; kurtosis = 12.43; BPD-related features at age 15, skewness = −1.75;
kurtosis = 3.70), we also ran the models using a robust maximum likelihood
estimator (MLR) that is less biased when facing departures from normality in
the data (Savalei, 2010; Yuan & Bentler, 2000). All results were similar to those
obtained using FIML, original model: χ2(36) = 179.10, p < .001; CFI = .91,
TLI = .88, RMSEA = .054; trimmed model: χ2(22) = 90.40, p < .001; CFI = .95,
TLI = .93, RMSEA = .05; delta CFI = .04. Individual effect sizes in the model
were also similar. Thus, the reported results were robust to these departures

G4797_480.indd 14 6/8/2020 3:27:32 PM

 DEVELOPMENTAL PATHWAYS TO BPD FEATURES 15

from normality. Because MLR does not model missing data, however, FIML
was retained as the primary estimator.

ASSESSMENT OF MEDIATING PATHS

In the final model in Figure 1, significant infancy/preschool predictors were
linked to BPDRF-15 by seven potential mediating paths (see Table 3). It should
be noted that most of these paths did not represent links across three differ-
ent developmental time points, so the results should not be interpreted as
implying etiological effects. However, the assessment of whether maternal
behavior in a later period was statistically mediating a link between maternal
behavior in an earlier period and child effects in a later period adds important
information to the model. Therefore, the significance of the seven potential
indirect paths was assessed using the confidence interval for the mediated
effects method. This method creates asymmetrical confidence intervals based
on the assumption of a nonnormal distribution of the indirect product term.
Evidence for mediation is assumed if the 95% confidence interval does not
include zero (Tofighi & MacKinnon, 2011). Six of the seven potential paths
were significant (Table 3).
These mediated effects fell into three groups. Two paths indicated that
early maternal insensitivity and early maternal depression were linked to
SEDR-MC in part through the continuity of maternal insensitivity and depres-
sion into middle childhood. Second, two paths indicated that maternal insen-
sitivity and depression in middle childhood were related to BPDRF-15 in part
through the emergence of social and emotional dysregulation in middle child-
hood. The final two paths indicated that maternal insensitivity and depression
in middle childhood were related to BPDRF-15 in part through the continuity
of maternal insensitivity and depression to age 15.

TABLE 3. Proposed Mediated Paths From Earlier Predictors to

Later Outcomes Through Mediating Variables
Indirect effect
Mediated Paths estimate (SE) CI
1. Infancy/preschool maternal sensitivity Maternal sensitivity middle −.023 (.006) [−.034, −.012]
childhood  SEDR-MC
2. Infancy/preschool maternal sensitivity  SEDR-MC BPDRF-15 −.004 (.002) [−.009, 0]
3. Maternal sensitivity middle childhood SEDR-MC BPDRF-15 −.008(.002) [−.012, −.004]
4. Maternal sensitivity middle childhood  Maternal sensitivity age −.009 (.003). [−.015, −.003]
15 BPDRF-15
5. Infancy/preschool maternal depression Maternal depression .013 (.002). [.01, .016]
middle childhood SEDR-MC
6. Maternal depression middle childhood SEDR-MC BPDRF-15 .004 (.001) [.003, .006]
7. Maternal depression middle childhoodMaternal depression at age .005 (.002) [.002, .008]
15 BPDRF-15
Note. SEDR-MC = social-emotional dysregulation in middle childhood; BPDRF-15 = BPD-related features at age 15.
Pathways with confirmed mediation are in bold. A mediated pathway from infancy/preschool maternal depression
through SEDR-MC to BPDRF-15 was not tested because the direct path from infancy/preschool maternal depression to
SEDR-MC was not significant.

G4797_480.indd 15 6/8/2020 3:27:32 PM

 16 BRUMARIU ET AL.

ALTERNATIVE PATH MODEL: INSENSITIVITY AS

MEDIATOR OF EFFECTS OF MATERNAL DEPRESSION?
In the current data, maternal depression was related to maternal sensitivity
both in early childhood (standardized estimate = −.31, p < .001) and in middle
childhood (standardized estimate = −.10, p < .001). The relation did not reach
significance at 15 years (standardized estimate = −.05, p = .13). Thus, we
also tested an alternative model in which the effects of maternal depression
were not independent but rather were mediated through maternal sensitivity.
To test this model, we removed the direct paths from maternal depression
to SEDR-MC and BPDRF-15 and added indirect paths through maternal
sensitivity to SEDR-MC and BPDRF-15. This mediated model provided a
relatively poor fit to the data, χ2(23) = 515.79, p < .001, CFI = .80, IFI = .80,
TLI = .61, and RMSEA = .13. Thus, results indicated that maternal depression
was not primarily affecting SEDR-MC and BPDRF-15 through decreases in
maternal sensitivity.

DO PARTICULAR CLASSES OF PARENTING DISTURBANCE

PREDICT SEDR-MC OR BPDRF-15?
Results of the path modeling indicated that pathways from maternal insensi-
tivity to adolescent BPDRF-15 had their beginnings in the infancy/preschool
period. Using the sensitivity composite from the SECCYD as a single continu-
ous variable was parsimonious for the path modeling and provided a consistent
metric for maternal behavior over all ages of the study. However, given the
importance of maternal behavior beginning in the infancy/preschool period
for later prediction of both SEDR-MC and BPDRF-15, and given the diverse
parenting problems that have been associated with these difficulties in the
literature, it was also important to generate a more fine-grained look at the
aspects of maternal behavior most predictive of later outcomes. Thus, an LCA
was conducted including the six rating scales indexing quality of mother–child
interaction at 54 months of age.
The LCA was run as an iterative model, starting with a single group and
continuing until the model fit was no longer better and statistically different
than the previous model. Model fit was assessed for one-, two-, three-, four-,
and five-group models. Based on the high entropy score (the closer to 1, the bet-
ter the fit) and the significant LMR, which is used to determine whether there
is statistically significant improvement over a model with one fewer classes
(k–1) (Nylund, Asparouhov, & Muthén, 2007), the four-class model provided
the best fit to the data (BIC = 21882.03, Entropy = .90, LMR = p < .01; see
Figure 2).
Descriptively, examination of the profiles in each class indicated the fol-
lowing characterizations: Class 1 was labeled “Optimal” and indexed sensitive,
responsive, and involved parenting, as shown by the high ratings on all positive
scales and low ratings on negative scales. This class included 36.76% of parents
(n = 386). Class 2, called “Sub-Optimal Support” indexed a parenting style in
which hostility remained low, but mean ratings for supportive presence and
respect for autonomy were approximately 1 scale point lower on average than

G4797_480.indd 16 6/8/2020 3:27:32 PM

 DEVELOPMENTAL PATHWAYS TO BPD FEATURES 17

FIGURE 2. Latent class analysis of 54-month observational ratings

of mother-child interaction.

the Optimal class, and scales for cognitive stimulation, quality of assistance,
and confidence were approximately 1.5 scale points lower. The Sub-Optimal
Support parenting group accounted for 47.23% of parents (n = 446). Moth-
ers in Class 3 were labeled “Withdrawn” because, while hostility remained
low, the mother’s lack of supportive involvement was notable. Support for
autonomy was 1.5 scale points lower than Optimal, but other rating items
were even lower, with supportive presence and confidence approximately
2.5 scale points lower than in the Optimal group, and quality of assistance
and cognitive stimulation approximately 3 scale points lower than in the
group labeled Optimal parenting. The Withdrawn parenting group accounted
for 15.04% of parents (n = 158). Class 4, labeled “Hostile-Unsupportive,”
made up the smallest group, accounting for only 4.76 % of parents (n = 50).
Mothers in the Hostile-Unsupportive group had scale scores for hostility that
were 3 scale points above all other groups. In addition, their scale scores for
respect for autonomy were lower than all other groups, indicating an intrusive
or interfering quality to the mother’s behavior. Furthermore, their scale scores
for supportive presence, quality of assistance, and confidence were similar to
those of the withdrawn group. Thus, mothers in this group combined high
levels of hostility and low levels of respect for autonomy with low levels of
assistance, stimulation, and support.

G4797_480.indd 17 6/8/2020 3:27:32 PM

 18 BRUMARIU ET AL.

Two one-way ANOVAs were then conducted using the four-class parent-
ing categorization as the independent variable and SEDR-MC and BPDRF-15
as the outcome variables. For middle childhood, prediction of SEDR-MC from
the four parenting classes was significant, F(3, 979) = 10.64, p < .001, η2 = .03,
and a post hoc Tukey analysis revealed that SEDR-MC was significantly
elevated among children whose mothers were in the Sub-Optimal, Withdrawn,
and Hostile parenting classes, compared to children whose mothers were in
the Optimal parenting class: all ps < .001, M (SD) optimal parenting = −.11
(.38), M (SD) suboptimal parenting = .02 (.48), M (SD) withdrawn parent-
ing = .09 (.60), M (SD) hostile parenting = .19 (.56).
For BPDRF-15, prediction from the four classes was also significant,
F(3, 888) = 4.90, p < .01, η2 = .02 (see Figure 3). However, post hoc Tukey
analyses revealed that only adolescents of mothers in the Withdrawn class
at 54 months had elevated BPDRF-15 compared to those of mothers in the
Optimal parenting class, M (SD)s = .09 (.58) and −.09 (.48), respectively,
p < .01. There was also a marginally significant difference between adolescents
whose mothers were in the Sub-Optimal parenting group, compared to those
whose mothers were in the Optimal parenting group, M (SD)s = .01 (.55) and
−.09 (.48), p =.059. No other comparisons were significant, M (SD) hostile
parenting = .12 (.63).

DISCUSSION

A number of investigators have shown that a coherent clustering of BPD-

related features can be identified in midadolescence (Sharp et al., 2014; Stepp
et al., 2014; Vaillancourt et al., 2014). In addition, the public health costs of
such behaviors are known to be considerable (e.g., Youngblade et al., 2006).
However, we still know relatively little about the earlier developmental path-
ways leading to BPD-related features in adolescence. The general develop-
mental model reviewed earlier posits that BPD-related features are rooted
in family processes that affect early social-emotional regulation, rather than
emerging in adolescence in response to new developmental pressures. Further,
the negative effect of poor parental regulation, combined with child tempera-
mental vulnerability, impedes the development of emotional regulation ability,
so that by middle childhood difficulties in emotion regulation also lead to
disturbed peer relationships. These difficulties in middle childhood further link
early risk factors to the development of BPD features in adolescence. Thus,
the first aim of the current study was to assess this general model and evaluate
the direct and indirect paths through which earlier risk factors influenced the
emergence of BPD-related features at age 15.
First, in bivariate correlations, maternal insensitivity during each develop-
mental period was significantly related to both social-emotional dysregulation
in middle childhood and to BPD-related features at age 15. Furthermore, in
the multivariate path model, insensitivity during each developmental period
continued to add independent prediction to the model with other variables
controlled. Thus, it mattered how long the child experienced maternal insensi-
tivity from infancy to adolescence in the prediction of BPDRF-15. In addition,

G4797_480.indd 18 6/8/2020 3:27:32 PM

 DEVELOPMENTAL PATHWAYS TO BPD FEATURES 19

mediation analyses indicated that maternal insensitivity in the infancy/pre-

school period contributed to BPDRF-15 primarily through the continuation
of maternal insensitivity into the grade school years. Insensitivity in middle
childhood then further contributed to BPDRF-15 by contributing to the emer-
gence of social and emotional dysregulation in middle childhood. In addition,
maternal insensitivity in middle childhood signaled a risk for continuity of
insensitivity into adolescence, which further added to prediction of BPDRF-
15. The model, then, tracks the repeated additional contributions of maternal
insensitivity at each developmental period from infancy to adolescence. Given
the strong continuity in maternal insensitivity over time shown in Figure 1,
these results suggest that efforts to prevent the emergence of both SEDR-MC
and BPDRF-15 need to begin with efforts to improve maternal sensitivity in
early childhood.
Maternal depression also emerged as a contributing factor to SEDR-MC
and to BPDRF-15, consistent with previous work (Belsky et al., 2012; Ham-
merton et al., 2016; Wolke et al., 2012). However, few studies have modeled
the contribution of maternal depression from infancy to adolescence, par-
ticularly in relation to BPD-related features. Results indicated, first, that there
was strong continuity in maternal depression across the three developmental
periods (Figure 1). Second, early maternal depression increased SEDR-MC
through the continuity of maternal depression into middle childhood. This
higher level of SEDR-MC further acted to increase the level of BPD-related
features at age 15. Finally, maternal depression in middle childhood also
influenced the level of BPD-related features at age 15 through the continuity
of maternal depression into adolescence. Notably, an alternative path model
assessing whether effects of maternal depression on BPDRF-15 were mediated
by maternal insensitivity did not provide a good fit to the data. Thus, in future
studies, other mediators of the effect of maternal depression on adolescent
BPD features will be important to explore, including the effects of shared
genes (Belsky et al., 2012).
Consistent with prior longitudinal reports (Carlson et al., 2009; Lyons-
Ruth et al., 2013), children who exhibited attachment disorganization in
infancy were more likely to exhibit increased BPD-related features in adoles-
cence. In addition, bivariate correlations indicated that disorganized/control-
ling forms of attachment at 36 months were particularly relevant to pathways
toward SEDR-MC, characterized by peer conflict, victimization, anger, and
suicidal thoughts. Notably, however, once maternal factors were entered into
the model, neither attachment assessment made an additional contribution to
prediction. Thus, while early disorganized and controlling forms of attachment
were part of a set of related predictors, maternal insensitivity and depression
over time better accounted for the variance in SEDR-MC and BPDRF-15.
The relative advantage of maternal assessments may be due to the repeated
assessment of maternal factors in the SECCYD, while child attachment was
assessed only up to 36 months of age. Since the SECCYD was conducted,
assessments of controlling attachment have also been validated in middle
childhood and adolescence (Brumariu et al., 2018; Bureau, Easterbrooks, &
Lyons-Ruth, 2009; Obsuth, Hennighausen, Brumariu, & Lyons-Ruth, 2014).
As noted earlier, using these more recent assessments, Lyons-Ruth et al. (2013)

G4797_480.indd 19 6/8/2020 3:27:32 PM

 20 BRUMARIU ET AL.

found that controlling attachment in middle childhood was a significant pre-

dictor of BPD features at age 19. Furthermore, Kobak et al. (2017) found
that controlling/caregiving interaction with parents at 13 years of age was a
significant predictor of increased risky behavior at 15 years. Finally, Lyons-
Ruth et al. (2015) found that controlling-caregiving interaction with parents
at age 19 was a concurrent correlate of both overall BPD features and sui-
cidality. Thus, future studies are needed that assess disorganized/controlling
forms of attachment in both middle childhood and adolescence in relation to
bully-victimization, conflict with peers, intense anger, suicidal ideation, and
risky behaviors.
Notably, consistent with the general developmental model, social-­
emotional dysregulation across Grades 3, 4, 5, and 6, including intense anger,
peer conflict, peer victimization, and suicidal thoughts, was predictive of BPD-
related features at age 15. Furthermore, SEDR-MC was an important mediator
of the effects of both maternal (in)sensitivity and maternal depression on age
15 outcome. These findings are important because BPDRF-15 is characterized
by a variety of risky behaviors not prominent in middle childhood, including
substance abuse, unsafe sexual behavior, risky driving, and disordered eat-
ing, as well as increased self-injury and suicidality. Thus, it is important to
identify precursors to these adolescent risky behaviors that can be assessed
in the grade school years. Our findings suggest that suicidal thoughts, anger,
peer conflict, and peer victimization are relevant aspects of dysregulation in
middle childhood that are linked to the more pervasive forms of risky self-
damaging behavior seen in adolescence. Therefore, preventive efforts in middle
childhood might significantly decrease the incidence of BPD-related features
in adolescence, with their serious public health consequences.
Finally, results of the person-centered LCA were notable in indicating
that all less-optimal parenting classes were associated with increased SEDR
in middle childhood, with hostile-unsupportive parenting showing the larg-
est effect. This finding is consistent with the evidence reviewed earlier of the
importance of harsh parenting on middle-childhood self-regulation, as well
as poor peer relations. In contrast, in relation to BPDRF-15, only early with-
drawn parenting was a significant predictor. Notably, a similar pattern of
results emerged in the only other longitudinal study to separately analyze both
hostile and withdrawing forms of early maternal behavior. Hostile maternal
interaction was particularly predictive of social-emotional maladaptation
in middle childhood (e.g., Lyons-Ruth, Easterbrooks, & Cibelli, 1997), but
withdrawing forms of maternal interaction were more salient in predicting
the risky, self-damaging features of BPD at age 19 (Lyons-Ruth et al., 2013).
The emergence of a similar pattern of results in the much larger SECCYD
sample points to the important role of active parental support and regulation
in the early years as a foundation for the child’s ability to effectively self-
regulate. In addition, the current data indicate that the effects of withdrawn
parenting on BPD-related features become particularly salient in adolescence
compared to middle childhood, perhaps because the adolescent increasingly
needs to self-regulate behavior in such important areas as sexuality, substance
use, eating behavior, self-care, driving, and handling money. While the conver-
gent results of these two studies point to the significant long-term effects of

G4797_480.indd 20 6/8/2020 3:27:32 PM

 DEVELOPMENTAL PATHWAYS TO BPD FEATURES 21

withdrawal during the preschool period, further work is needed to separately

assess withdrawal and hostility as parenting components in middle childhood
and adolescence, as well as to evaluate how these different forms of maladap-
tive parenting might affect different facets of development over time. Notably,
withdrawn patterns of maternal interaction have not been studied as exten-
sively as parental hostility/harsh discipline, so that assessments of maternal
withdrawal should be included in future work on developmental contributors
to the dysregulations in affect and behavior characteristic of BPD.
The latent class results were also important in indicating that withdraw-
ing patterns of maternal behavior may be more prevalent than hostile patterns
of maternal behavior. Only 5% of the sample was classified in the Hostile
Unsupportive group at 54 months, while 15%, or three times as many, were
classified in the Withdrawn category. Finally, another 47% were rated as
suboptimal in support and assistance provided to the child, and even those
smaller decrements in support were associated with a significant increase in
social-emotional dysregulation in middle childhood. Thus, if we are to inter-
vene effectively to prevent the risky self-damaging behaviors characteristic
of BPD, withdrawing patterns of maternal behavior, marked by low levels of
supportive presence, quality of assistance and cognitive stimulation, may need
increased attention in intervention efforts.
While the SECCYD was unparalleled in its repeated observations of
parent–­child interaction and its rich assessment of maladaptive child behaviors
over time, limitations of the study should also be noted. First, the SECCYD
did not include a validated measure of BPD features in adolescence because
the study was conducted before such validated assessments were available.
Thus, our measure of BPD-related features relied on items from other validated
assessments of related constructs. In addition, assessments of maltreatment
were not included, so the contributions of maltreatment to social-emotional
dysregulation and BPD-related features could not be evaluated (Zanarini et al.,
1997). Furthermore, family psychiatric assessments were not conducted. Belsky
et al. (2012) found that the link between exposure to harsh treatment through
age 10 and emotional/behavioral dysregulation at age 12 was stronger among
children with a family history of psychiatric illness. Thus, parental psychiatric
assessments would be important to include in future work and might add to
the current findings regarding the role of maternal depression. In addition, the
effect of maternal depression on child dysregulation over time was indepen-
dent of maternal sensitivity, suggesting a role for other mediators, including
shared genetic factors, which could be explored in future work. Relatedly, the
low-risk nature of the SECCYD sample may have underestimated the influ-
ence of some predictors, in that more extreme values may be more frequent
in high-risk samples. Thus, these effects need to be replicated in more diverse
samples with a higher risk of psychopathology.
Notably also, the interactions of fathers and children were not consis-
tently assessed, so the role of fathers in these pathways remains an important
topic for future work. As noted in the Results section, we assessed indirect
effects that were based, in part, on cross-sectional associations rather than
on three distinct temporal waves. Thus, etiological inferences should not be
drawn. Replication of these mediated paths is needed using nonconcurrent

G4797_480.indd 21 6/8/2020 3:27:32 PM

 22 BRUMARIU ET AL.

assessments in a multiwave design. In addition, randomized intervention

designs will be needed to fully assess etiological effects. Finally, we evaluated
a relatively parsimonious model from a theoretical standpoint which posits
that maternal insensitivity and depression in earlier periods contribute to later
child and adolescent BPD-related features. However, bidirectional effects are
equally important to test because it is likely that children exhibiting increased
BPD-related features may affect mothers’ ability to remain sensitive to their
children’s needs and may negatively affect mothers’ mood and capacity to
self-regulate.
In summary, the findings point to a long developmental trajectory toward
BPD-related features in adolescence, with early maternal insensitivity and
depression likely to continue into middle childhood and contribute to middle
childhood social-emotional dysregulation. This increased dysregulation in mid-
dle childhood then increases the BPD-related features emerging in adolescence.
In addition, the continuation of maternal depression and insensitivity into
adolescence further increases the level of BPD-related features in adolescence.
Fortunately, these long-term developmental trajectories also point to
several potential targets for intervention. First, there are now a number of
evidence-based interventions for increasing early maternal sensitivity and
decreasing disorganized attachment (for a review, see Facompré, Bernard,
& Waters, 2018). The current results also suggest that such efforts should
address maternal withdrawal, in particular, to increase active maternal sup-
port, regulation, and stimulation for the child (Yarger, Bronfman, Carlson,
& Dozier, 2019). Second, school-based interventions are also indicated by
the current findings to prevent the emergence of the disturbed peer relations
and emotional dysregulation that forecast increased BPD features in adoles-
cence. Finally, working with adolescents and their parents to improve sensitive
parent–­adolescent communication is indicated as an important intervention
target, as is increased treatment of maternal depression.

REFERENCES
Achenbach, T. M. (1991a). Manual for the Child
Behavior Checklist/ 4-18 and 1991 Pro-
file. Burlington, VT: University of Vermont,
Department of Psychiatry.
Achenbach, T. (1991b). Manual for the Youth Self-
Report and 1991 Profile. Burlington, VT: Uni-
versity of Vermont, Department of Psychiatry.
Ainsworth, M. D. S., Bell, S. M., & Stayton, D. J.
(1974). Infant-mother attachment and social
development: Socialisation as a product of
reciprocal sensitivity to signals. In M. P. M.
Richards (Ed.), Introduction of the child into
a social world (pp. 99–135). Cambridge, UK:
Cambridge University Press.
Ainsworth, M. D. S., Blehar, M. C., Waters, E., &
Wall, S. (1978). Patterns of attachment: A
psychological study of the strange situation.
Oxford, England: Lawrence Erlbaum.
Alati, R., Baker, P., Betts, K. S., Connor, J. P., Lit-
tle, K., Sanson, A., & Olsson, C. A. (2014).
The role of parental alcohol use, parental
discipline and antisocial behaviour on ado-
lescent drinking trajectories. Drug & Alcohol
Dependence, 13, 178–184. https://doi.org/10
.1016/j.drugalcdep.2013.09.030
American Psychiatric Association. (2013). Diagnos-
tic and statistical manual of mental disorders
(5th ed.). Arlington, VA: American Psychiat-
ric Publishing.
Belsky, D. W., Caspi, A., Arseneault, L., Bleidorn, W.,
Fonagy, P., Goodman, M., . . . Moffitt, T. E.
(2012). Etiological features of borderline
personality related characteristics in a birth
cohort of 12-year-old children. Development
and Psychopathology, 24, 251–265. https://doi
.org/10.1017/s0954579411000812

G4797_480.indd 22 6/8/2020 3:27:32 PM

 DEVELOPMENTAL PATHWAYS TO BPD FEATURES 23
Brener, N., & Collins, J. (1998). Co-occurrence of
health-risk behaviors among adolescents
in the United States. Journal of Adolescent
Health, 22, 209–213. https://doi.org/10
.1016/s1054-139x(97)00161-4
Brumariu, L.E., Giuseppone, K.R., Kerns, K.A.,
Van de Walle, M., Bureau, J-F., Bosmans,
G., & Lyons-Ruth, K. (2018). Middle child-
hood attachment strategies: Validation of
an observational measure. Attachment &
Human Development, 20, 491–513. https://
doi.org/10 .1080/14616734.2018.1433696
Bureau, J. F., Easterbrooks, M. A., & Lyons-Ruth, K.
(2009). Attachment disorganization and
controlling behavior in middle childhood:
Maternal and child precursors and correlates.
Attachment & Human Development, 11,
265–284. https://doi.org/10.1080/14616730
902814788
Campbell, S., Morgan-Lopez, A., Cox, M., &
McLoyd, V. C. (2009). A latent class analy-
sis of maternal depressive symptoms over 12
years and offspring adjustment in adoles-
cence. Journal of Abnormal Psychology, 118,
479–493. https://doi.org/10.1037/a0015923
Carlson, E. A., Egeland, B., & Sroufe, L. A. (2009).
A prospective investigation of the devel-
opment of borderline personality symp-
toms. Development and Psychopathology,
21, 1311–1334. https://doi.org/10.1017/
s0954579409990174
Cassidy, J., Marvin, R. S., & MacArthur Working
Group. (1992). Attachment organization in
three- and four-year-olds: Procedures and
coding manual. Unpublished manual, Uni-
versity of Maryland.
Centers for Disease Control and Prevention.
(2017). Web-based Injury Statistics Query
and Reporting System (WISQARS). Atlanta,
GA: National Center for Injury Prevention
and Control. Retrieved from https://www.cdc
.gov/injury/wisqars/leading_causes_death
.html
Chanen, A., & Kaess, M. (2012). Developmental
pathways to borderline personality disorder.
Current Psychiatry Reports, 14(1), 45–53.
https://doi.org/10.1007/s11920-011-0242-y
Clark, K. E., & Ladd, G. W. (2000). Connectedness
and autonomy support in parent-child rela-
tionships: Links to children’s socioemotional
orientation and peer relationships. Develop-
mental Psychology, 36, 485–498. https://doi
.org/10.1037//0012-1649.36.4.485
Cole, P. M., Llera, S. J., & Pemberton, C. K. (2009).
Emotional instability, poor emotional aware-
ness, and the development of borderline per-
sonality. Development and Psychopathology,
21, 1293–1310. https://doi.org/10.1017/
S0954579409990162
Conger, R. D., & Rand, G. H. (1994). Families in
troubled times: Adapting to change in rural
America. New York, NY: A. de Gruyter.
G4797_480.indd 23
Crowell, S. E., Beauchaine, T. P., & Linehan, M. M.
(2009). A biosocial developmental model
of borderline personality: Elaborating and
extending Linehan’s theory. Psychological
Bulletin, 135, 495–510. https://doi.org/10
.1037/a0015616
DuRant, R., Smith, J., Kreiter, S., & Krowchuk, D.
(1999). The relationship between early age of
onset of initial substance use and engaging in
multiple health risk behaviors among young
adolescents. Archives of Pediatrics & Ado-
lescent Medicine, 153, 286–291. https://doi
.org/10.1001/archpedi.153.3.286
Earnshaw, V. A., Elliott, M. N., Reisner, S. L., Mrug, S.,
Windle, M., Emery, S. T., . . . Schuster, M. A.
(2017). Peer victimization, depressive symp-
toms, and substance use: A longitudinal anal-
ysis. Pediatrics, 139, e20163426. https://doi
.org/10.1542/peds.2016-3426
Facompré, C., Bernard, K., & Waters, T. E.
(2018). Effectiveness of interventions in
preventing disorganized attachment: A
meta-analysis. Development and Psychopa-
thology, 30, 1–11. https://doi.org/10.1017/
s0954579417000426
Fonagy, P., & Luyten, P. (2009). A developmen-
tal, mentalization-based approach to the
understanding and treatment of borderline
personality disorder. Development and Psycho-
pathology, 21, 1355–1381. https://doi.org/10
.1017/S0954579409990198
Garner, D. M., Olmsted, M. P., Bohr, Y., & Garfin-
kel, P. E. (1982). The Eating Attitudes Test:
Psychometric features and clinical correlates.
Psychological Medicine, 12, 871–878. https://
doi.org/10.1017/s0033291700049163
Goodman, S. H., & Gotlib, I. H. (Eds.). (2002).
Children of depressed parents: Alterna-
tive pathways to risk for psychopathology.
Washington, DC: American Psychological
Association.
Gore, F. M., Bloem, P., Patton, G. C., Ferguson, J.,
Joseph, V., Coffrey, C., . . . Mathers, C. (2011).
Global burden of disease in young people
aged 10–24 years: A systematic analysis. The
Lancet, 377(9783), 2093–2102. https://doi
.org/10.1016/s0140-6736(11)60512-6
Guilamo-Ramos, V., Bouris, A., Lee, J., McCar-
thy,  K., Michael, S. L., Pitt-Barnes, S., &
Dittus, P. (2012). Paternal influences on
adolescent sexual risk behaviors: A struc-
tured literature review. Pediatrics, 130(5),
e1313–e1325. https://doi.org/10.1542/peds
.2011-2066
Gunderson, J., & Lyons-Ruth, K. (2008). BPD’s
interpersonal hypersensitivity phenotype:
A gene-environment transactional model.
Journal of Personality Disorders, 22, 22–41.
https://doi.org/10.1521/pedi.2008.22.1.22
Hammerton, G., Zammit, S., Thapar, A., & Col-
lishaw, S. (2016). Explaining risk for suicidal
ideation in adolescent offspring of mothers
6/8/2020 3:27:32 PM
 24 BRUMARIU ET AL.
with depression. Psychological Medicine, 46,
265–275. https://doi.org/10.1017/s00332
91715001671
Klomek, A. B., Sourander, A., Niemelä, S., Kumpu-
lainen, K., Piha, J., Tamminen, T., . . . Gould,
M. S. (2009). Childhood bullying behaviors
as a risk for suicide attempts and completed
suicides: A population-based birth cohort
study. Journal of the American Academy of
Child & Adolescent Psychiatry, 48, 254–
261. https://doi.org/10.1097/chi.0b013e
318196b91f
Kobak, R., Zajac, K., Abbott, C., Zisk, A., & Bou-
noua, N. (2017). Atypical dimensions of
caregiver–adolescent interaction in an eco-
nomically disadvantaged sample. Develop-
ment and Psychopathology, 29, 405–416.
https://doi.org/10.1017/S0954579417
000074
Kochenderfer, B. J., & Ladd, G. W. (1996). Peer
victimization: Cause or consequence of
school maladjustment? Child Development,
67, 1305–1317. https://doi.org/10.2307/
1131701
Lee, K. S., & Vaillancourt T. (2018). Developmental
pathways between peer victimization, psy-
chological functioning, disordered eating
behavior, and body mass index: A review and
theoretical model. Aggression and Violent
Behavior, 39, 15–24. https://doi.org/10.1016
/j.avb.2018.01.004
Lereya, S. T., Winsper, C., Heron, J., Lewis, G., Gun-
nell, D., Fisher, H. L., & Wolke, D. (2013).
Being bullied during childhood and the pro-
spective pathways to self-harm in late adoles-
cence. Journal of the American Academy of
Child & Adolescent Psychiatry, 52, 608–618.
https://doi.org/10.1016/j.jaac.2013.03.012
Linehan, M. (1993). Cognitive-behavioral treat-
ment of borderline personality disorder. New
York, NY: Guilford Press.
Lyons-Ruth, K., Brumariu, L. E., Bureau, J. F., Hen-
nighausen, K., & Holmes, B. (2015). Role
confusion and disorientation in young adult-
parent interaction among individuals with
borderline symptomatology. Journal of Per-
sonality Disorders, 29, 641–662. https://doi
.org/:10.1521/pedi_2014_28_165
Lyons-Ruth, K., Bureau, J. F., Holmes, B., Easter­
brooks, A., & Brooks, N. H. (2013). Bor-
derline symptoms and suicidality/self-injury
in late adolescence: Prospectively observed
relationship correlates in infancy and
childhood. Psychiatry Research, 206(2–3),
273–281. https://doi.org/10.1016/j.psychres
.2012.09.030
Lyons-Ruth, K., Easterbrooks, M. A., & Cibelli,
C. D. (1997). Infant attachment strategies,
infant mental lag, and maternal depressive
symptoms: Predictors of internalizing and
externalizing problems at age 7. Develop-
G4797_480.indd 24
mental Psychology, 33, 681–692. https://doi
.org/10.1037/0012-1649.33.4.681
Lyons-Ruth, K., Wolfe, R., Lyubchik, A., Gro-
gan, T., & Steingard, R. (2002). Prevalence
and correlates of depressive symptoms in
parents of children under three: Results of
the Commonwealth Survey. In N. Halfon,
K. McLearn, & M. Schuster (Eds.), Child
rearing in America: Challenges facing par-
ents with young children (pp. 217–259).
Cambridge, UK: Cambridge University Press.
Main, M., & Cassidy, J. (1988). Categories of
response to reunion with the parent at age 6:
Predictable from infant attachment classifica-
tions and stable over a 1-month period. Devel-
opmental Psychology, 24, 415–426. https://
doi.org/10.1037/0012-1649.24.3.415
Main, M., & Solomon, J. (1990). Procedures for
identifying disorganized/disoriented infants
during the Ainsworth Strange Situation. In
M. T. Greenberg, D. Cicchetti, & E. M. Cum-
mings (Eds.), Attachment in the preschool
years (pp. 121–160). Chicago, IL: University
of Chicago Press.
Meade, A. W., Johnson, E. C., & Braddy, P. W.
(2008). Power and sensitivity of alternative
fit indices in tests of measurement invariance.
Journal of Applied Psychology, 93, 568–592.
https://doi.org/10.1037/0021-9010.93.3.568
Michonski, J. D., Sharp, C., Steinberg, L., &
Zanarini, M. C. (2013). An item response
theory analysis of the DSM-IV borderline
personality disorder criteria in a population-
based sample of 11- to 12-year-old children.
Personality Disorders: Theory, Research,
and Treatment, 4, 15–22. https://doi.org/
10.1037/a0027948
NICHD Early Child Care Research Network.
(2001). Nonmaternal care and family fac-
tors in early development: An overview of the
NICHD Study of Early Child Care. Journal
of Applied Developmental Psychology, 22,
457–492. https://doi.org/10.1016/s0193
-3973(01)00092-2
Nylund, K. L., Asparouhov, T., & Muthén, B. O.
(2007). Deciding on the number of classes
in latent class analysis and growth mix-
ture modeling: A Monte Carlo simula-
tion study. Structural Equation Modeling,
14, 535–569. https://doi.org/10.1080
/10705510701575396
Obsuth, I., Hennighausen, K., Brumariu, L. E., &
Lyons-Ruth, K. (2014). Disorganized behav-
ior in adolescent–parent interaction: Relations
to attachment state of mind, partner abuse,
and psychopathology. Child Development,
85, 370–387. https://doi.org/10.1111/cdev
.12113
Pena, J. B., Matthieu, M. M., Zayas, L. H., Masyn,
K. E., & Caine, E. D. (2012). Co-occurring
risk behaviors among White, Black, and
6/8/2020 3:27:32 PM
 DEVELOPMENTAL PATHWAYS TO BPD FEATURES 25
Hispanic US high school adolescents with
suicide attempts requiring medical attention,
1999–2007. Social Psychiatry and Psychiatric
Epidemiology, 47, 29–42. https://doi.org/10
.1007/s00127-010-0322-z
Peterson, B. S., Zhang, H., Santa Lucia, R., King,
R. A., & Lewis, M. (1996). Risk factors
for presenting problems in child psychiat-
ric emergencies. Journal of the American
Academy of Child & Adolescent Psychia-
try, 35, 1162–1173. https://doi.org/10
.1097/00004583-199609000-00014
Radloff, L. S. (1977). The CES-D Scale: A self-
report depression scale for research in the
general population. Applied Psychological
Measurement, 1, 385–401. https://doi.org/10
.1177/014662167700100306
Savalei, V. (2010). Small sample statistics for
incomplete nonnormal data: Extensions of
complete data formulae and a Monte Carlo
comparison. Structural Equation Model-
ing, 17, 241–264. https://doi.org/10.1080
/10705511003659375
Schafer, J., & Graham, J. W. (2002). Missing data:
Our view of the state of the art. Psychologi-
cal Methods, 7, 147–177. https://doi.org/10
.1037/1082-989x.7.2.147
Sharp, C., Steinberg, L., Temple, J., & Newlin, E.
(2014). An 11-item measure to assess bor-
derline traits in adolescents: Refinement of
the BPFSC using IRT. Personality Disorders:
Theory, Research, and Treatment, 5, 70–78.
https://doi.org/10.1037/per0000057
Stepp, S. D., Lazarus, S. A., & Byrd, A. L. (2016).
A systematic review of risk factors prospec-
tively associated with borderline personality
disorder: Taking stock and moving forward.
Personality Disorders, 7, 316–323. https://
doi.org/10.1037/per0000186
Stepp, S. D., Whalen, D. J., Pilkonis, P. A., Hipwell,
A. E., & Levine, M. D. (2011). Children of
mothers with borderline personality dis-
order: Identifying parenting behaviors as
potential targets for intervention. Personality
Disorders: Theory, Research, and Treatment,
3, 76–91. https://doi.org/10.1037/a0023081
Stepp, S. D., Whalen, D. J., Scott, L. N., Zalewski, M.,
Loeber, R., & Hipwell, A. E. (2014). Recipro-
cal effects of parenting and borderline per-
sonality disorder symptoms in adolescent
girls. Development & Psychopathology, 26,
361–378. doi:10.1017/S0954579413001041
Tafà, M., Cimino, S., Ballarotto, G., Bracaglia,
F., Bottone, C., & Cerniglia, L. (2017).
Female adolescents with eating disorders,
parental psychopathological risk and fam-
ily functioning. Journal of Child & Family
Studies, 26, 28–39. https://doi.org/10.1007
/s10826-016-0531-5
Tofighi, D., & MacKinnon, D. (2011). R mediation: An
R package for mediation analysis confidence
G4797_480.indd 25
interval. Behavior Research Methods, 43,
692–700. https://doi.org/10.3758/s13428
-011-0076-x
Vaillancourt, T., Brittain, H., McDougall, P., Krygs-
man, A., Boylan, K., Duku, E., & Hymel, S.
(2014). Predicting borderline personality dis-
order symptoms in adolescents from child-
hood physical and relational aggression,
depression, and attention-deficit/hyperactiv-
ity disorder. Development and Psychopathol-
ogy, 26, 817–830. https://doi.org/10.1017/
S0954579414000418
van IJzendoorn, M. H., Schuengel, C., & Bakermans-
Kranenburg, M. J. (1999). Disorganized
attachment in early childhood: Meta-
analysis of precursors, concomitants, and
sequelae. Development and Psychopathol-
ogy, 11, 225–250. https://doi.org/10.1017
/s0954579499002035
Weinberger, D. A., & Schwartz, G. E. (1990). Distress
and restraint as superordinate dimensions of
self-reported adjustment: A typological per-
spective. Journal of Personality, 58, 381–417.
https://doi.org/10.1111/j.1467-6494.1990
.tb00235.x
Winsper, C. (2018). The aetiology of borderline
personality disorder (BPD): Contemporary
theories and putative mechanisms. Current
Opinion in Psychology, 21, 105–110. https://
doi.org/10.1016/j.copsyc.2017.10.005
Winsper, C., Lereya, S. T., Marwaha, S., Thomp-
son, A., Eyden, J., & Singh, S. P. (2016). The
aetiological and psychopathological validity
of borderline personality disorder in youth:
A systematic review and meta-analysis. Clini-
cal Psychology Review, 44, 13–24. https://doi
.org/10.1016/j.cpr.2015.12.001
Winsper, C., Zanarini, M., & Wolke, D. (2012).
Prospective study of family adversity and
maladaptive parenting in childhood and bor-
derline personality disorder symptoms in a
non-clinical population at 11 years. Psycho-
logical Medicine, 42, 2405–2420. https://doi
.org/10.1017/s0033291712000542
Wolke, D., Schreier, A., Zanarini, M. C., & Win-
sper, C. (2012). Bullied by peers in childhood
and borderline personality symptoms at 11
years of age: A prospective study. Journal
of Child Psychology &Psychiatry, 53, 846–
855. https://doi.org/10.1111/j.14697610
.2012.02542.x
Yarger, H., Bronfman, E., Carlson, E. & Dozier,
M. (2019). Intervening with attachment
and biobehavioral catch-up to decrease dis-
rupted parenting behavior and attachment
disorganization: The role of parental with-
drawal. Development and Psychopathology.
Advance online publication. https://doi.
org/10.1017/S0954579419000786
Youngblade, L. M., Curry, L. A., Novak, M., Vogel,
B., & Shenkman, E. A. (2006). The impact
6/8/2020 3:27:32 PM
 26 BRUMARIU ET AL.
of community risks and resources on adoles-
cent risky behavior and health care expen-
ditures. Journal of Adolescent Health, 38,
486–494. https://doi.org/10.1016/j.jado-
health .2005.07.016
Yuan, K. H., & Bentler, P. M. (2000). Robust mean
and covariance structure analysis through
iteratively reweighted least squares. Psy-
chometrika, 65(1), 43–58. https://doi.org/10
.1007/BF02294185
Zanarini, M. C., Barison, L. K., Frankenburg, F. R.,
Reich, D. B., & Hudson, J. I. (2009). Family
history study of the familial coaggregation
G4797_480.indd 26
of borderline personality disorder with axis
I and nonborderline dramatic cluster axis II
disorders. Journal of Personality Disorders,
23, 357–369. https://doi.org/:10.1521/pedi
.2009.23.4.357
Zanarini, M. C., Williams, A. A., Lewis, R. E., Reich,
R. B., Vera, S. C., Marino, M. F., . . . Fran-
kenburg, F. R. (1997). Reported pathological
childhood experiences associated with the
development of borderline personality dis-
order. American Journal of Psychiatry, 154,
1101–1106. https://doi.org/10.1176/ajp
.154.8.1101
6/8/2020 3:27:32 PM