Development and Psychopathology 31 (2019), 173–187

# Cambridge University Press 2019

doi:10.1017/S0954579418001530

Protective factors that buffer against the intergenerational

transmission of trauma from mothers to young children:
A replication study of angels in the nursery

ANGELA J. NARAYAN,a,b CHANDRA GHOSH IPPEN,b WILLIAM W. HARRIS,c AND ALICIA F. LIEBERMANb
a
Department of Psychology, University of Denver; b Department of Psychiatry and Child Trauma Research Program, University of
California, San Francisco; and c Children’s Research and Education Institute, New York City

Abstract
This replication study examined protective effects of positive childhood memories with caregivers (“angels in the nursery”) against lifespan and
intergenerational transmission of trauma. More positive, elaborated angel memories were hypothesized to buffer associations between mothers’ childhood
maltreatment and their adulthood posttraumatic stress disorder (PTSD) and depression symptoms, comorbid psychopathology, and children’s trauma exposure.
Participants were 185 mothers (M age ¼ 30.67 years, SD ¼ 6.44, range ¼ 17–46 years, 54.6% Latina, 17.8% White, 10.3% African American, 17.3% other;
24% Spanish speaking) and children (M age ¼ 42.51 months; SD ¼ 15.95, range ¼ 3–72 months; 51.4% male). Mothers completed the Angels in the Nursery
Interview (Van Horn, Lieberman, & Harris, 2008), and assessments of childhood maltreatment, adulthood psychopathology, children’s trauma exposure,
and demographics. Angel memories significantly moderated associations between maltreatment and PTSD (but not depression) symptoms, comorbid
psychopathology, and children’s trauma exposure. For mothers with less positive, elaborated angel memories, higher levels of maltreatment predicted higher
levels of psychopathology and children’s trauma exposure. For mothers with more positive, elaborated memories, however, predictive associations were not
significant, reflecting protective effects. Furthermore, protective effects against children’s trauma exposure were significant only for female children,
suggesting that angel memories may specifically buffer against intergenerational trauma from mothers to daughters.
Keywords: childhood maltreatment; intergenerational transmission; protective factors; resilience

Decades of research have focused on the intergenerational Very little research has identified protective factors within
transmission of stress and adversity from parents to children, the parent that are not contingent exclusively on the contribu-
elucidating risk factors that propagate the cycle of maltreat- tions of others. This study focuses on “angels in the nursery,”
ment, violence, and other forms of trauma within families defined as individuals’ capacity to recall memories of loving
(Ertem, Leventhal, & Dobbs, 2000; Sroufe, Egeland, Carl- moments with their caregivers. Although angel memories
son, & Collins, 2005; Widom & Wilson, 2015). There is may stem at least in part from actual interpersonal experi-
much less research on protective factors that buffer against ences, the qualities that characterize how they are recalled
the transmission of trauma from the childhoods of parents (e.g., the clarity, coherence, richness, and specificity of the
to the childhoods of their offspring. Research on protective descriptions) are believed to vary as a function of individual
factors that alleviate parenting disturbances and break the cy- differences and the unique meaning and salience of the mem-
cle of abuse mostly focuses on parents’ interpersonal re- ories for the specific individual. Individuals who can hold
sources, such as support from their romantic partners and onto benevolent childhood memories when they become par-
therapeutic interventions that promote “meaning-making” ents may be able to draw internal resources from these mem-
of past abuse and foster positive parent–child relationships ories and re-create positive caregiving moments with their
(Egeland & Eriksen, 2004; Egeland, Jacobvitz, & Sroufe, offspring as buffers against the intergenerational transmission
1988; Geeraert, van den Noortgate, Grietens, & Onghena, of adversity (Lieberman, Padrón, Van Horn, & Harris, 2005).
2004; Lieberman & Van Horn, 2008; Raby, Steele, Carlson, The theoretical basis for these premises is rooted in the litera-
& Sroufe, 2015; Roisman, Padrón, Sroufe, & Egeland, 2002). ture on psychological defenses elaborated by psychoanalytic
infant mental health contributions (Fraiberg, 1980) and
This study honors the memory of Patricia Van Horn, who was an original de- attachment theory (Bowlby, 1980).
veloper of the Angels in the Nursery Interview. We also thank the participat-
ing mothers who shared their angel memories with us, and their children who
took part in this study. Background of Angels in the Nursery
Address correspondence and reprint requests to: Angela Narayan, Uni-
versity of Denver, Department of Psychology, 2155 S. Race St., Denver, The concept of “angels in the nursery” emerged as a counter-
CO 80208; E-mail: Angela.Narayan@du.edu. part to the metaphor of “ghosts in the nursery” (Fraiberg,
173
174 A. J. Narayan et al.

Adelson, & Shapiro, 1975), which describes the conflicted af- with childhood caregivers. Rather than internal working
fects stemming from a parent’s unresolved childhood experi- models of attachment that shape relational expectations, how-
ences of impaired caregiving that are suppressed and later ever, angel memories are specific recollections of concrete
reenacted with the parent’s offspring. These “ghosts in the moments characterized by love, tenderness, and protection.
nursery” are often elicited by the young offspring’s displays For instance, although the generalized feeling of being loved
of distress, fear, anger, or other negative affects that trigger in the latter example above may be considered as dismissive
the parents’ often unconscious recollections of their suppres- or idealized on the Adult Attachment Interview because it
sed or compartmentalized early affective experiences. Alter- lacks supporting details, it is given a score in the moderate
natively, the term “angels in the nursery” describes a parent’s range on the Angels Interview because it denotes an internal
memories of loving childhood experiences with a caregiver state where defenses against affect do not interfere with an or-
that may serve as guiding models for nurturing the parent’s ganized state of mind that involves feeling loved (Bowlby,
offspring (Lieberman et al., 2005). These benevolent memo- 1988).
ries may coexist with memories of fear and distress in relation While many parents are readily able to recall angel mem-
to childhood caregivers and may persist into adulthood in ories following the opening Angels Interview prompt, some
spite of childhood adversities and unresolved trauma. parents who experienced childhood adversity are not able
The first evidence of “angels in the nursery” originated in to recall any memory of feeling loved with a caregiver. Pilot
clinical self-descriptions of mothers receiving child–parent findings (Narayan et al., 2017) showed that in a sample of 54
psychotherapy (CPP) with their young children following mothers reared in foster care, approximately 15% were unable
experiences of intimate partner violence. Several of the to recall any angel memories and received scores of 0 for an-
mothers reported that they were able to maintain positive gel memory quality (see Table 1 for examples of all possible
parenting because they remembered childhood experiences scores). Of note, angel memory scores were not associated
of love and support that sustained them despite simultaneous with the severity of childhood maltreatment in that sample,
or subsequent experiences of violence and victimization suggesting that access to more positive, elaborated angel
(Lieberman et al., 2005; Lieberman & Van Horn, 2008). memories is not merely a function of having experienced
Clinical case studies also illustrate that remembering angel less severe childhood maltreatment.
memories enables traumatized parents to bring moments of Although it is likely that having frequent and consistent
love, compassion, and safety to their children, resulting in childhood experiences of positive caregiving may provide a
therapeutic benefits (Narayan, Oliver Bucio, Rivera, & respondent with a greater reservoir of angel memories to
Lieberman, 2016). draw from, it is often the case that respondents describe richly
“Angel memories” are assessed using the Angels in the elaborated angel memories despite also reporting childhoods
Nursery Interview (abbreviated here to “Angels Interview”), characterized by predominantly chronic and severe adversity.
which prompts respondents to “think of a memory of a time The role of psychic defenses in how individuals organize
when you were little when you felt especially loved, under- their experience of reality has been extensively discussed in
stood, or safe” (Van Horn, Lieberman, & Harris, 2008; Nara- psychoanalytic theory and attachment theory, both of which
yan, Ghosh Ippen, Harris, & Lieberman, 2017). Memory nar- share a common root in acknowledging the ubiquity of
ratives are audio-recorded, transcribed, and later coded by psychological conflict as the result of incompatibilities
trained raters for positive valence, elaboration, richness, and between competing motivations and resulting thoughts and
specificity (Ghosh Ippen, Narayan, Van Horn, & Lieberman, affects (Fonagy, Gergely, & Target, 2008). It is possible
2015). An example of a positive and elaborated angel mem- that some individuals are able to derive great comfort and
ory (i.e., one that is specific and richly detailed) receiving a long-term benefit from even fleeting or inconsistent
high score of 4 out of 5 is “My grandfather always made experiences of feeling loved and protected, whereas other
me feel understood because he was so patient. I had a lot individuals fend off positive memories because their sense
of imaginary friends, and he humored me. One time when of self and their perception of their caregivers are organized
we were driving, I was worried about my imaginary friend, around their experiences of stressful and traumatic interper-
and I said, ‘Joey forgot his shoe.’ My grandfather turned sonal events.
the car right around to retrieve the lost imaginary shoe.” A Memory is fluid, and angel memories that are initially inac-
slightly lower score of 3, which is the most frequent score, re- cessible due to defensive processes may emerge in the course
flects memories of generally feeling loved without providing of therapeutic work that supports an exploration of the emo-
specific examples, such as, “I always felt loved when I was tional impact of adversity and trauma. CPP has been helpful
little. I don’t have an exact memory because the feeling was in facilitating recall of positive and elaborated angel memories
always there.” in mothers who were not able to recall any angel memories
Angel memories are conceptualized as independent but re- during the initial assessment before starting psychotherapy
lated constructs to other well-validated assessments of posi- (Narayan et al., 2016). Resolving and healing from childhood
tive early life experiences. For instance, the Adult Attachment adversity during the course of CPP or other trauma-informed
Interview (Main & Goldwyn, 1984a, 1984b) assesses inter- interventions may provide access to angel memories by un-
nalized representations of attachment security or insecurity blocking defensive processes such as deactivating (dismissing
Intergenerational protective effects of angels in the nursery 175

Table 1. Examples of angel memories ranging in scores from 0 to 5

Description Score Example

No positive memory 0 “I can’t really remember anything about my childhood. I don’t have a good memory,
so I find I usually can’t recall things from the past.”
0 “I can’t think of anything because I don’t remember ever feeling that way.”
0 “No, I have no good memories of when I was little. What I do have sometimes are
flashbacks of being abused.”
Vague, hollow, or shallow 1 “I remember being at my uncle’s house a lot. We watched TV. It was fun.”
1 “In school. Hanging out with my older sister, and her friends, who were older than
me. We ate together, laughed. I liked being out of my house.”
1 “When I went to Mexico with my grandmother for two weeks and I got to know my
extended family.”
Isolated, transient, or fleeting 2 “I remember everything was fine, nothing special. The word ‘love’ wasn’t really
used, although every year on my birthday, my parents took me on a special outing,
like to the zoo or to a museum, and went with me to do whatever I wanted.”
2 “One year I remember having a childcare teacher who really liked me. She took me
under her wing, and when my mom didn’t take me to daycare, she would call to
ask where I was. I didn’t stay at that school very long because we moved.”
2 “One time when I was sick, everyone came to visit me, including my dad who
worked far away.”
Global, general, or nonspecific 3 “My grandma would always cook my favorite food, which was an authentic Peruvian
dish, when I would go to her house on the weekends. It made me feel really
special.”
3 “I don’t have specific memories, but I remember my dad used to hug me a lot. He
also snuck me candies, cookies, things like that.”
3 “My mom always supported me. I trusted her and could tell her anything.”
Elaborated, specific, and detailed with a 4 “Bedtime was good. I remember my parents would take turns reading and singing to
sensory experience and/or a quality of me (while the other one was with my sister). Dad would sing ‘Home on the
comfort, protection, reassurance, etc., Range,’ and Mom would sing ‘Rock-a-bye, Baby.’ I remember feeling warm and
after a negative feeling (e.g., worried, cuddly, with all my stuffed animals around me.”
scared, or hurt)
4 “My mom gave me a lot of attention and physical affection, like kisses, because I was
prone to getting hurt. One time I remember I fell off my bike and scraped my face,
and my mom rubbed my head and held me in her lap.”
4 “I would go to my aunt’s house after nursery school when my mom was still
working. My aunt always kept a satin pillow and special blanket for me to use
during my nap. I remember it smelled like almond extract, and it made me feel
safe.”
4 “My grandmother was very busy, but I remember one day she bought me a tea set. I
sat between her legs so she could braid my hair to get ready for the ‘tea party.’
Then, I set up the tea set under the table, and she crawled under to play with me.”
Multiple elaborated, specific, and 5 At least two memories that independently qualify as “4.”
detailed

or avoidant) attachment strategies and by segregating mental childhood maltreatment and their adulthood posttraumatic
systems to exclude painful affective memories from con- stress disorder (PTSD) symptoms. The first aim of the present
sciousness (Bowlby, 1980). study was to determine whether angel memories also buffered
against the association between mothers’ childhood maltreat-
ment and their current PTSD symptoms in a larger, multieth-
The Current Study nic, multilingual sample. We hypothesized that mothers’
angel memories would moderate the association between
The present study aimed to replicate and extend pilot evi- their childhood maltreatment and their total PTSD symptoms,
dence from a study that empirically tested angels in the nur- consistent with protective effects.
sery as protective factors that buffer against the intergenera- The second aim was to examine whether angel memories
tional transmission of trauma in high-risk families (Narayan also buffered against the association between childhood mal-
et al., 2017). In the pilot study of 54 mothers reared in child- treatment and other forms of adulthood psychopathology
hood foster care, more positive and elaborated angel memo- because childhood trauma is typically associated with psychi-
ries were buffers against the association between mothers’ atric comorbidities (Putnam, Harris, & Putnam, 2013). We
176 A. J. Narayan et al.

examined additional forms of psychopathology, including public hospital following their young children’s (ages 0–6
depression symptoms and comorbid psychopathology (de- years) exposure to a traumatic event (e.g., exposure to family
fined as clinical levels of PTSD and depression symptoms). or community violence, direct child maltreatment, serious ill-
We expected that more positive and elaborated angel memo- ness, incarceration of a family member, death of someone
ries would also buffer against the association between close, etc.). The child for whom the mother was seeking ser-
mothers’ childhood maltreatment and their depression vices was the target child in this study (51.4% male, M ¼
symptoms and comorbid psychopathology. 42.51 months, SD ¼ 15.95, range ¼ 3–72 months; 50.8% La-
The third aim was to examine the protective effects of an- tina, 18.9% biracial/multiracial, 14.1% White, 10.3% African
gel memories in the link between mothers’ childhood mal- American, 4.3% Asian, 1.6% other). Families were referred to
treatment and their children’s trauma exposure. The pilot the study by sources including but not limited to mental health
study did not include the participants’ children, precluding service providers, case managers, child welfare workers, pri-
the ability to ascertain the intergenerational protective effects mary care providers, other clients, and self-referrals.
of angel memories on the parents’ offspring. In the current Prior to receiving CPP, mothers and children participated
study, we hypothesized that mothers’ more positive and in an extensive pretreatment assessment, which served as the
elaborated angel memories would moderate the association source of data for the current study. Mothers completed stan-
between their childhood maltreatment and their children’s dardized instruments on life events, including their childhood
exposure to traumatic events, again reflecting protective maltreatment, current psychopathology symptoms (i.e.,
effects. PTSD and depression), their children’s lifetime exposure to
The fourth and final exploratory aim was to examine traumatic events, demographic and contextual information
whether significant moderation of mothers’ angel memories for covariate variables, and the Angels Interview, which
in the link between their childhood maltreatment and their was audio-recorded and later coded by trained raters. Trained
children’s trauma exposure differed by the child’s sex. Pre- master’s- or doctoral-level clinical researchers conducted all
vious research indicates that there are sex differences in the assessments in mothers’ preferred language of English or
types of trauma transmitted to male versus female offspring Spanish. The institutional review board at the University of
of parents who were maltreated as children. For instance, California, San Francisco, approved all study procedures,
the intergenerational transmission of childhood sexual abuse and mothers provided informed consent for themselves and
is particularly apparent from mothers to daughters (McClos- their children prior to enrolling. Mothers received honoraria
key & Bailey, 2000). Mothers who were maltreated as chil- to compensate them for completing the assessment, and
dren also show sex-specific abusive and punitive parenting children received small toys.
of their offspring, such that mother–son relationships tend
to include higher levels of boundary violations in physical in-
Measures: Primary variables
timacy, whereas mother–daughter relationships often include
higher levels of derisiveness and belittling (Shaffer & Sroufe, Angel memories. Mothers completed the Angels in the Nursery
2005). Despite these sex-specific patterns of risk transmis- Interview, which consists of seven standardized questions that
sion, it is not known whether sex-specific patterns of protec- ask participants to think of “a memory of a time when you
tion also exist. Exploring sex differences was included as a were little when you felt especially loved, understood, or
goal of the study because of the dearth of research on this safe.” Follow-up prompts probe about the age of the memory,
aspect of intergenerational resilience. sensory details (e.g., smells, sights, sounds, or other sensations
connected with the memory), and additional memories with the
same or different childhood caregivers or attachment figures
Method
that also elicit feelings of being loved, understood, or safe
(Narayan et al., 2017; Van Horn et al., 2008). Thus, respondents
Participants
are given several opportunities to generate memories, and their
The study participants were 185 biological mothers (M age ¼ final angel memory score reflects the memory that is the most
30.67 years, SD ¼ 6.44, range ¼ 17–46 years) who self-iden- positive and elaborated. For example, a mother who described
tified as Latina (54.6%), White (17.8%), African American two memories, one qualifying as a “3” and one qualifying as a
(10.3%), biracial/multiracial (9.2%), Asian (5.4%) or other “4,” would receive a final score of 4.
(2.7%). Almost one quarter of the mothers (23.8%) were The Angels Interview typically requires 10–15 min to ad-
monolingual Spanish speaking. Mothers reported an average minister and comparable time to code, following approxi-
educational attainment of less than a high school degree (M mately 8 hr of training. Trained coders, one of whom was
¼ 11.28 years, SD ¼ 2.70, range ¼ 1–13 years) and predomi- bilingual and a native Spanish speaker, reviewed the memory
nantly low-income status (M average monthly income ¼ narratives and assigned scores on a 6-point scale for total
$1,873.30, SD ¼ $2,192.19, range ¼ $0–$16,000), with memory quality. According to the coding manual (Ghosh
75% of the sample reporting an average income of $2,000 Ippen et al., 2015), higher scores prototypically reflect angel
per month. Mothers were referred for clinical services (CPP) memories that are more positive, specific, elaborated, and
at a university-affiliated clinical research program at an urban detailed, and include the presence of either (a) a sensory
Intergenerational protective effects of angels in the nursery 177

experience that adds to the quality of the memory (e.g., being Maternal comorbid psychopathology. A comorbid psychopa-
hugged or held, smelling a scent such as perfume or food as- thology score was calculated by determining whether
sociated with a loving caregiver), or (b) reassuring feelings, mothers met criteria for clinical levels of PTSD and depres-
such as comfort or protection, in the context of feeling sad, sion diagnoses according to the DTS and the BDI-2, respec-
lost, threatened, or some other negative emotion. (See Table 1 tively. For clinical PTSD, DTS scores were dichotomized at
for descriptions of angel memories and examples of responses 40 to reflect the “efficiency cutoff” recommended by the
receiving each score.) Interrater reliability for angel memories developers (Davidson et al., 1997). This score of 40 corre-
was excellent (intraclass correlation coefficient ¼ .92). sponds to the score at which participants in the original valid-
ity study were diagnosed with PTSD based on the Structured
Mothers’ childhood maltreatment. Mothers completed the Clinical Interview for the DSM-III-R (Spitzer, Williams, Gib-
Life Stressors Checklist—Revised (Wolfe, Kimerling, bon, & First, 1990). For clinical depression, BDI-2 scores
Brown, Chrestman, & Levin, 1996), composed of 30 adver- were dichotomized at 10 to reflect at least mild to moderate
sities, such as childhood maltreatment, as well as noninter- depression, based on recommendations from developers
personal stressors (e.g., experiences of natural disasters, car (Beck, Steer, & Carbin, 1988). Dichotomized clinical-level
accidents, physical illness, etc.). The present study relied on PTSD and depression scores were added, yielding a comorbid
maternal endorsement of five maltreatment subtypes occur- psychopathology score: 0 (neither clinical PTSD nor depres-
ring between ages 0 and 17: emotional abuse/neglect (being sion, n ¼ 35; 18.9%), 1 (either clinical PTSD or depression,
“frequently shamed, embarrassed, ignored, or repeatedly told n ¼ 49; 26.5%), or 2 (both, n ¼ 77; 41.6%).
you were ‘no good’”; n ¼ 83, 44.9%); physical neglect (being
“not fed, properly clothed, or left to take care of oneself when Children’s exposure to traumatic events. Mothers completed
too young or ill”; n ¼ 46, 24.9%); exposure to interparental the Traumatic Events Screening Inventory—Parent Form,
violence, often considered a form of emotional abuse (Holt, Revised (Ghosh Ippen et al., 2002), which assesses child
Buckley, & Whelan, 2008; seeing “violence between family exposure to 23 different lifetime traumatic events including
members, e.g., hitting, kicking, slapping, and punching”; n ¼ child maltreatment, exposure to family and community
102, 55.1%); physical abuse (being abused or physically violence, accidental injury, incarceration or loss/separation
attacked, e.g., hit, slapped, choked, burned, or beat up”; from caregivers, and other adversities. A sum of total trau-
n ¼ 86, 46.5%); or sexual abuse (having ever been “touched matic event exposure was computed (M ¼ 5.71, SD ¼ 2.61,
or made to touch someone else in a sexual way because he/ range ¼ 0–15).
she forced you in some way or threatened to harm you if
you didn’t” or had “sex [oral, anal, genital] when you didn’t
want to because someone forced you in some way or threa- Measures: Covariates
tened to harm you if you didn’t”; n ¼ 99, 53.5%). Scores Demographics. Several demographic covariates were gath-
on maternal maltreatment ranged from 0 to 5 (M ¼ 2.27, ered from the maternal assessment and used in the analyses,
SD ¼ 1.68). For analyses, these five items were summed including mothers’ preferred language (English or Spanish),
for total experiences of childhood maltreatment. maternal age, average monthly income, and educational
attainment; and child age and sex. There was a wide distribu-
Mothers’ PTSD symptoms. Mothers completed the Davidson tion of maternal educational attainment from first grade
Trauma Scale (DTS; Davidson et al., 1997), a standardized through postsecondary education. Given the skewness in
instrument that assesses 17 self-reported PTSD symptoms this variable, educational attainment was trichotomized into
in the past week. Items are rated on 0- to 4-point frequency three groups: less than a high school education (n ¼ 51;
scales ranging from not at all to every day (total frequency 27.6%), high school degree or equivalent (n ¼ 42; 22.7%),
scores range from 0 to 68) and severity scales from not at and at least some college (n ¼ 84; 47.6%).
all distressing to extremely distressing (total severity scores
range from 0 to 68). According to standard procedures, fre- Maternal stress. Given that mothers were the primary infor-
quency and severity scales are combined to yield a total raw mants on all variables except for coder-rated angel memories,
score ranging from 0 to 136 for total PTSD symptoms (M maternal stress at the time of the assessment was controlled
¼ 47.89, SD ¼ 29.94; range ¼ 0–125; a ¼ 0.94). using the parental distress subscale of the 36-item Parenting
Stress Index, Short Form, Version 3 (Abidin, 1990). This sub-
Maternal depression symptoms. Mothers completed the Beck scale assesses general parental stress in the parenting role on
Depression Inventory, Second Edition (BDI-2; Beck, Steer, 0- to 5-point scales from strongly agree to strongly disagree
& Brown, 1996), which assesses 21 self-reported depression (M ¼ 31.18, SD ¼ 10.19, range ¼ 12–60; a ¼ 0.89). Raw
symptoms in the past 2 weeks on 0- to 3-point frequency scores were used as covariates. (Of note, raw scores 31
scales. Total scores range from 0 to 63. Analyses used the correspond to 85th percentile according to the developers
raw score for total depression symptoms (M ¼ 17.17, SD ¼ and reflect substantially elevated parental stress, which was
11.45, range ¼ 0–50; a ¼ 0.92). present in 42.7% [n ¼ 79] of this sample.)
178 A. J. Narayan et al.

Data analytic plan ferences in the pattern of findings were noted; thus, all of the
results reported below reflect the raw data.
All hypotheses were tested using hierarchical linear regres-
sions with interaction terms, yielding four total regressions:
maternal PTSD symptoms, depression symptoms, and co- Results
morbid psychopathology; and children’s exposure to trau-
matic events. Spanish language was considered a theoretical Descriptive statistics
covariate and included as a control variable in all analyses be-
Rates of mothers’ childhood maltreatment were high, with
cause this is the first empirical study to examine the protective
149 mothers (80.5%) reporting at least one subtype of
effects of angel memories in multilingual families. Thus, it
childhood maltreatment, and 24 mothers (13.0%) reporting
was important to examine whether protective effects held
all five subtypes. Table 2 presents bivariate correlations,
after accounting for potential associations between maternal
means, standard deviations (SD), and sample ranges for all
language and maternal/child outcomes. Maternal stress was
study variables. Higher levels of maternal childhood maltreat-
also considered a theoretical covariate and was included in
ment were significantly associated with higher levels of
all regressions to observe whether findings held after account-
PTSD symptoms, (r ¼ .21, p , .01), depression symptoms
ing for mothers’ stress levels because they provided reports
(r ¼ .21, p , .01), and comorbid psychopathology (r ¼
on all of the other variables. Given the overlap between
.30, p , .01), and were marginally significantly associated
maternal stress and the three psychopathology outcomes (par-
with higher levels of children’s trauma exposure (r ¼ .14,
ticularly depression symptoms), all analyses were conducted
p , .10). However, higher levels of childhood maltreatment
with and without maternal stress as a covariate, and all results
were not significantly associated with less positive angel
remained the same regardless. The results below are
memories (r ¼ –.05, p ¼ ns).
presented with maternal stress included.
In terms of empirical covariates, higher maternal educa-
Additional covariates (e.g., maternal age, educational
tional attainment was marginally significantly associated
attainment, and average monthly income; and child age and
with lower maternal depression symptoms (r ¼ –.15, p ,
sex) were considered to be empirical covariates (i.e., poten-
.10) and significantly associated with higher offspring trauma
tially sample-specific and not conceptually presumed to nec-
exposure (r ¼ .31, p , .01), so educational attainment was in-
essarily relate to outcomes). These covariates were included in
cluded in regressions predicting to those outcomes. Higher ma-
analyses when they were significantly associated with any of
ternal average monthly income was marginally significantly
the maternal or child outcomes at the bivariate level. For each
associated with lower maternal depression symptoms (r ¼
regression, significant covariates were entered in the first step,
–.15, p , .10), so income was also included in the regression
main effects (i.e., maternal childhood maltreatment and angel
predicting to depression symptoms. Maternal age was not sig-
memories, z-scored) were entered in the second step, and the
nificantly associated with any outcome, so it was not included
interaction of maternal maltreatment and angel memories
in the regression analyses. Older child age was significantly
was entered in the third step. All regression results were exam-
positively correlated with greater child trauma exposure (r ¼
ined for any influential cases according to Cook’s d  4/n
.26, p , .01), and child male sex was marginally negatively as-
(Cook & Weisberg, 1982; Rawlings, 1988). No influential
sociated with child trauma exposure (r ¼ –.13, p , .10), so
cases significantly affected any of the results reported below.
both variables were included in that regression.
Significant interactions were probed with the Johnson–Ney-
man technique (Hayes & Mattes, 2009) to examine significant
simple slopes of the outcome on maternal childhood maltreat- Regression analyses
ment at different values of the moderator, angel memories.
Maternal PTSD symptoms. Higher levels of maternal child-
hood maltreatment significantly predicted higher levels of
mothers’ PTSD symptoms, revealing a main effect of mal-
Missing data
treatment (b ¼ .20, p , .05). There was a significant interac-
Missing data were minimal and ranged from 0% on most tion of MaltreatmentAngel Memories (b ¼ –.15, p ¼ .05)
demographic covariates, angel memories, and children’s in predicting PTSD symptoms (Table 3). (The p value for this
trauma exposure; to approximately 2%–4% on maternal interaction dropped below .05 according to the imputed re-
stress, average monthly income, and educational attainment; sults, the results without influential cases, and the results
to 13.0% on certain maternal outcomes (e.g., comorbid psy- without maternal stress controlled). The interaction of Mal-
chopathology). The total percentage of missing data across treatment  Angel Memories independently added a small
the entire data set was only 3.6%. Data were assumed to be but significant amount of variance in the final step of the
missing at random. Twenty data sets were imputed with fully model, DR2 ¼ .02, p ¼ .05. In terms of covariates, Spanish
conditional specification in SPSS Version 24.0 (Rubin, 1987; language marginally predicted lower PTSD symptoms (b ¼
Schafer & Graham, 2002), and all regression analyses were –.13, p , .10), and higher levels of maternal stress signifi-
reconducted using results pooled across the 20 imputed data cantly predicted higher levels of PTSD symptoms (b ¼ .25,
sets. All raw and imputed results were compared, and no dif- p , .01). The final model predicted approximately 15% of
 Table 2. Bivariate correlations and descriptive statistics for all study variables

Primary variables 1 2 3 4 5 6 7 8 9 10 11 12 13

1. Maternal childhood maltreat. —

2. Angel memories –.05 —
3. Maternal PTSD sxs. .21** –.02 —
4. Maternal depression sxs. .21** –.12 .55** —
5. Maternal comorbid psychop. .30** –.02 .78** .71**
6. Children’s trauma exposure .14† .10 .37** .22** .33** —
Demographic/stress covariates
7. Maternal language: Spanish –.09 .13† –.14† –.05 –.08 –.10 —
8. Maternal stress .07 –.17* .26** .55** .37** .11 .01 —
9. Maternal educational attain. –.01 .26** .03 –.15† –.02 .31** –.29** –.21** —
10. Maternal age (years) –.03 .04 –.05 –.13 –.02 .06 –.11 –.25** .29** —
11. Average monthly income –.05 .02 –.10 –.15† –.09 .07 –.19** .20** .34** .32** —
12. Child age (months) –.01 .04 .07 –.03 .05 .26** –.05 .00 .14† .23* .14† —
13. Child sex (male) –.06 –.05 –.01 –.01 .02 –.13† .06 –.06 –.03 .00 .04 –.05 —
Descriptive statistics
Mean (or %) 2.27 3.04 47.89 17.17 1.26 5.71 23.7% 31.18 11.28 30.67 1,873.50 42.51 51.4%
SD 1.68 1.23 29.94 11.45 .80 2.61 NA 10.19 2.70 6.44 1,200 15.95 NA
Sample range 0–5 0–5 0–125 0–50 0–2 0–15 NA 12–60 1–13 17–46 0–16,000 3–72 NA

Note: Maltreat., maltreatment. Sxs., symptoms. Psychop., psychopathology. Attain., attainment. *p , .05. **p , .01. †p , .10.
179

Table 3. Hierarchical regression for maternal PTSD symptoms and comorbid psychopathology

Maternal PTSD symptoms Maternal comorbid psychopathology

B SE ß 95% CI R2 F DR2 B SE ß 95% CI R2 F DR2

Step 1: Covariates .10 5.70** .10** .15 9.34** .15**

Spanish language 210.57 5.35 –.15† 220.94 20.21 –.19 .14 –.10 –.46 .08
Maternal stress 0.76 0.22 .26** 0.32 1.19 .03 .01 .37** .02 .04
Step 2: Main effects .13 4.58** .03† .22 8.34** .07**
Spanish language 29.12 5.29 –.13† 219.57 1.31 –.14 .14 –.08 –.40 .13
Maternal stress 0.71 0.22 .25** 0.27 1.15 .03 .01 .35** .02 .04
Maternal childhood maltx 5.27 2.27 .18* 0.80 9.75 .20 .06 .25** .08 .31
Angel memories 0.41 2.25 .01 24.03 4.85 .02 .06 .03 –.09 .13
Step 3: Interaction .15 4.53** .02^ .24 7.76** .02*
Spanish language 28.69 5.24 –.13† 219.04 1.67 –.12 .13 –.07 –.39 .14
Maternal stress 0.71 0.22 .25** 0.28 1.15 .03 .01 35** .02 .04
Maternal childhood maltx 5.97 2.27 .20* 1.48 10.46 .21 .06 .27** .10 .33
Angel memories 1.38 2.28 .05 23.13 5.89 .05 .06 .06 –.07 .16
Maltx × Angel Memories 24.70 2.38 –.15^ 29.40 0.00 –.12 .06 –.15* –.24 –.01

Note: Maltx, maltreatment. *p , .05. **p , .01. ^p ¼ .05. †p , .10.

180 A. J. Narayan et al.

Figure 1. Angel memories buffer effects of maternal childhood maltreatment on (a) adulthood PTSD symptoms, (b) adulthood comorbid
psychopathology, and (c) children’s trauma exposure. Angel memories were tested as a continuous moderator, as described in text, but are
graphically presented at three levels of the moderator here for simplicity.

the variance in maternal PTSD symptoms. Probing the signif- of mothers’ depression symptoms, revealing a main effect
icant interaction (Figure 1a) revealed that the simple slopes of of maltreatment (b ¼ .15, p , .05), but the interaction of Mal-
maternal maltreatment and PTSD symptoms were significant treatment  Angel Memories was not significant (Table 4).
at 1 SD below the mean on angel memories through approxi- Higher maternal stress significantly predicted higher
mately the mean on angel memories (z scores ,.48), corre- levels of maternal depression symptoms (b ¼ .55, p ,
sponding to angel memory scores 3, but not at .48 SD .01). Together, the final model predicted approximately
above the mean on angel memories (corresponding to 35% of the variance in maternal depression symptoms. Given
scores of 4 or 5). In other words, childhood maltreatment the nonsignificant interaction, no further analyses were con-
significantly predicted adulthood PTSD symptoms only for ducted.
mothers with less positive, elaborated angel memories (who
received scores 3). Maternal comorbid psychopathology. Similar to the regres-
sion for maternal PTSD symptoms, higher levels of maternal
Maternal depression symptoms. Higher levels of maternal childhood maltreatment significantly predicted higher levels
childhood maltreatment significantly predicted higher levels of mothers’ comorbid psychopathology (b ¼ .27, p , .01),
Intergenerational protective effects of angels in the nursery 181

Table 4. Hierarchical regression for maternal depression symptoms

Maternal depression symptoms

B SE ß 95% CI R2 F DR2

Step 1: Covariates .33 13.76** .33**

Spanish language 0.59 1.97 –.07 25.68 2.12
Maternal stress 0.59 0.08 .54** 0.44 0.75
Maternal educational attainment 20.54 1.04 –.04 22.60 1.52
Average monthly income 0.00 0.00 –.03 0.00 0.00
Step 2: Main effects .35 10.64 .02
Spanish language 21.13 2.0 –.04 25.14 2.88
Maternal stress 0.59 0.08 .54** 0.43 0.74
Maternal educational attainment 20.38 1.08 –.03 22.52 1.76
Average monthly income 0.00 0.00 –.03 0.00 0.00
Maternal childhood maltreatment 1.63 0.77 .15* 0.10 3.16
Angel memories 20.05 2.03 –.04 0.43 0.75
Step 3: Interaction .35 9.43 .00
Spanish language 21.08 2.03 –.04 25.09 2.93
Maternal stress 0.59 0.08 .54** 0.43 .75
Maternal educational attainment 20.53 10.9 –.04 22.69 1.64
Average monthly income 0.00 0.00 –.02 0.00 .00
Maternal childhood maltreatment 1.72 0.78 .15* 0.18 3.26
Angel memories 0.15 0.84 .01 21.50 1.80
Maltreatment × Angel Memories 20.82 0.82 –.07 22.44 .81

Note: *p , .05. **p , .01.

revealing a main effect of maltreatment. There was a signifi- exposure, DR2 ¼ .03, p , .05. Significant covariates included
cant interaction of Maltreatment  Angel Memories (b ¼ higher levels of maternal stress (b ¼ .19, p , .05) and mater-
–.15, p , .05) in predicting comorbid psychopathology nal educational attainment (b ¼ .29, p , .01), and older child
(Table 3). This interaction also predicted a small but signifi- age (b ¼ .19, p , .01). The final model predicted approxi-
cant amount of variance in comorbid psychopathology as an mately 22% of the variance in children’s trauma exposure.
independent step, DR2 ¼ .02, p , .05. In terms of covariates, Probing the interaction (Figure 1c) revealed that similar to
higher levels of maternal stress also predicted higher risk for maternal PTSD symptoms, the simple slopes of maternal
comorbid psychopathology (b ¼ .35, p , .01). The final childhood maltreatment on their children’s trauma exposure
model predicted approximately 24% of the variance in were significant at 1 SD below the mean, through approxi-
comorbid psychopathology. Probing the interaction mately the mean on angel memories (z scores  –.03), corre-
(Figure 1b) showed that the simple slopes of maternal sponding to scores on angel memories 3, but not at ..–03
maltreatment on comorbid psychopathology were significant SD above the mean on angel memories (corresponding to
at 1 SD below the mean, through slightly above the mean scores of 4 or 5 on angel memories). Following a similar pat-
(z scores ,.85), corresponding to angel memory scores tern as PTSD symptoms, mothers’ childhood maltreatment
4, but not at .85 SD above the mean on angel memories significantly predicted their children’s exposure to more trau-
(corresponding to scores of 5 on angel memories). In other matic events only when mothers recalled less positive and
words, childhood maltreatment significantly predicted adult- elaborated angel memories (and received scores 3).
hood comorbid psychopathology unless mothers received the
highest possible score on angel memories. Post hoc analyses for child sex differences. Exploratory re-
gression analyses were conducted with the total sample split
Children’s traumatic event exposure. The regression for by child sex to understand whether protective effects of ma-
children’s exposure to traumatic events revealed similar ternal angel memories replicated in buffering sons’ versus
patterns as maternal PTSD symptoms and comorbid psycho- daughters’ exposure to greater numbers of traumatic events.
pathology: higher levels of maternal childhood maltreatment The regression predicting children’s trauma exposure was re-
significantly predicted higher levels of their children’s trauma conducted separately for male versus female children. Results
exposure (b ¼ .15, p , .05). Furthermore, the interaction be- showed that for males/sons, higher levels of maternal child-
tween Maltreatment  Angel Memories (b ¼ –.17, p , .05) hood maltreatment did not significantly predict higher levels
was also significant in predicting children’s trauma exposure of their trauma exposure. The interaction of Maltreatment 
(Table 5). This interaction effect independently contributed a Angel Memories in predicting male children’s trauma expo-
small but significant amount of variance to children’s trauma sure was also not significant.
182 A. J. Narayan et al.

Table 5. Hierarchical regression for offspring exposure to traumatic events

Children’s traumatic event exposure

B SE ß 95% CI R2 F DR2

Step 1: Covariates .18 5.86** .18**

Spanish language –.05 .47 –.01 20.99 0.88
Maternal stress .05 .02 .18* 0.01 0.08
Maternal educational attainment .94 .24 .31** 0.46 1.42
Child age (months) .03 .01 .19** 0.01 0.06
Child sex (male) –.54 .37 –.10 21.28 0.20
Step 2: Main effects .19 4.84** .01
Spanish language –.03 .49 .00 20.98 0.93
Maternal stress .05 .02 .18* 0.01 0.08
Maternal educational attainment .9 .15 .30** 0.42 1.41
Child age (months) .03 .01 .19** 0.01 0.06
Child sex (male) –.51 .37 –.10 21.25 0.23
Maternal childhood maltreatment .32 .19 .12† 20.05 0.69
Angel memories .12 .20 .05 20.27 0.51
Step 3: Interaction .22 5.04** .03*
Spanish language –.04 .48 –.01 20.99 0.90
Maternal stress .05 .02 .19* 0.01 0.08
Maternal educational attainment .87 .25 .29** 0.38 1.36
Child age (months) .03 .01 .19** 0.01 0.05
Child sex (male) –.45 .37 –.09 21.18 0.28
Maternal childhood maltreatment .39 .19 .15* 0.02 0.75
Angel memories .23 .20 .09 20.17 0.63
Maltreatment × Angel Memories –.46 .20 –.17* 20.85 20.07

Note: *p , .05. **p , .01. †p , .10.

For girls/daughters, higher levels of maternal childhood Discussion

maltreatment predicted higher levels of their trauma exposure
This study replicates and extends published pilot findings
(b ¼ .26, p , .01). The interaction of Maltreatment Angel
showing the protective role of benevolent childhood memo-
Memories was also significant (b ¼ –.24, p , .05), and it
ries with caregivers (“angel memories”) against lifespan
independently contributed a small but significant amount of
and intergenerational trauma. Findings from the pilot study
variance in the final step of the model, DR2 ¼ .05, p , .05.
(Narayan et al., 2017) revealed that for mothers who had
Higher maternal stress was also a significant predictor of
more positive and elaborated angel memories, the association
daughters’ higher levels of trauma exposure (b ¼ .28, p ,
between their childhood maltreatment and adult PTSD symp-
.01), as was older child age (b ¼ .24, p , .01). Mothers’
toms was not significant, consistent with protective effects.
higher educational attainment was a marginally significant
The present study replicated these findings. For mothers
predictor of daughters’ higher levels of trauma exposure.
with less positive and elaborated angel memories, childhood
(b ¼ .23, p , .10). The final model predicted approximately
maltreatment significantly predicted PTSD symptoms. Alter-
29% of the overall variance in girls’/daughters’ traumatic
natively, for mothers with more positive and elaborated angel
event exposure. Probing the interaction revealed a similar
memories, higher levels of their childhood maltreatment did
cutpoint for angel memories as effective buffers against
not predict higher levels of their PTSD symptoms, suggesting
daughters’ trauma exposure as the regression did with
that their angel memories protected them from trauma-related
both sexes together. The simple slopes of maternal child-
psychopathology in adulthood.
hood maltreatment on daughters’ trauma exposure were
significant at 1 SD below the mean, through approximately
the mean on angel memories (z scores ,.09), corresponding
Replication and extension of angel memories as protective
to scores on angel memories 3, but not at .09 SD above
factors
the mean on angel memories (corresponding to scores of
4 or 5 on angel memories). That is, higher levels of mothers’ The replicability of angel memories as protective factors that
childhood maltreatment significantly predicted daughters’ buffer against the effects of mothers’ childhood maltreatment
exposure to more traumatic events when mothers’ angel on their adult PTSD symptoms has now been established in
memories were less positive and elaborated (and received two different samples with unique ethnic, cultural, and lin-
scores 3). guistic differences, and with method variance. In the pilot
Intergenerational protective effects of angels in the nursery 183

sample, mothers were primarily African American, all were were more proximal predictors of depression symptoms. In
native English speakers, all had been reared for at least part support of this possibility, mothers’ stress, operationalized
of their childhoods in foster care, and many had experienced as general distress in the parenting role, was very strongly as-
very severe childhood maltreatment (Narayan et al., 2017). In sociated with their depression symptoms. Although maternal
the present study, the sample was more than three times as stress did not influence whether or not angel memories mod-
large and consisted of predominantly Latina mothers, with erated the association between childhood maltreatment and
approximately one-quarter of mothers identifying as mono- maternal depression symptoms (moderation was nonsignifi-
lingual Spanish speakers. In addition, the pilot study and cant with or without maternal stress included in the regression
the current study utilized different instruments to assess the model for depression), it is possible that contemporaneous
same independent and dependent variables, further generaliz- stressors, rather than childhood adversity or angel memories,
ing replicability across methods. The pilot study used the were among the strongest influences on depression symptoms
Childhood Trauma Questionnaire (Bernstein et al., 2003) in this group of mothers.
for mothers’ childhood maltreatment and the Trauma Symp- Furthermore, it is also possible that memories of positive
tom Inventory, Second Edition (Briere, Elliott, Harris, & childhood experiences are stronger buffers against PTSD
Cottman, 1995) for maternal PTSD symptoms, whereas the symptoms than depression symptoms in samples of maltreated
current study utilized the Life Stressors Checklist—Revised mothers. A similar pattern was observed in a separate sample
for maternal childhood maltreatment (Wolfe et al., 1996) of low-income pregnant women with high levels of childhood
and the Davidson Trauma Scale for PTSD symptoms (David- adversity: higher benevolent childhood experiences (BCEs;
son et al., 1997). Together, these findings provide strong and Narayan, Rivera, Bernstein, Harris, & Lieberman, 2018),
replicable evidence that more positive and elaborated angel which index the overall availability of childhood supports
memories can buffer maltreated mothers from PTSD symp- and resources rather than memories of them, significantly pre-
toms in adulthood. dicted lower PTSD but not depression symptoms after ac-
The present study also investigated the protective effects counting for women’s childhood adversity (Narayan et al.,
of maternal angel memories on additional aspects of maternal 2018). However, whether and to what extent higher levels of
and child well-being. Findings showed that more positive and angel memories are effective protective factors against depres-
elaborated angel memories buffered maltreated mothers from sion symptoms, as well as other sequelae following childhood
experiencing comorbid psychopathology in adulthood (oper- adversity, should continue to be examined in other samples.
ationalized as clinical levels of PTSD and depression symp- This is the first empirical study to document that mothers’
toms) and also buffered their offspring from being exposed angel memories buffer against the intergenerational transmis-
to higher numbers of traumatic events. Of note, however, sion of trauma from mothers to children. We found that for
protective effects of angel memories were not significant mothers with more positive and elaborated angel memories,
for maternal depression symptoms. These findings indicate the association between maternal childhood maltreatment
that the protective effects of angel memories may be specific and their children’s exposure to traumatic events was non-
to psychological outcomes reflecting traumatic stress (e.g., significant, reflecting a protective effect. Alternatively, for
PTSD symptoms and offspring trauma exposure) rather mothers with less positive and elaborated angel memories,
than to more general psychological outcomes. Furthermore, higher levels of mothers’ childhood maltreatment predicted
individuals with childhood maltreatment, especially that their children’s higher trauma exposure. One possible path-
which is chronic or co-occurring across subtypes, are often way may involve the negative impact of mothers’ childhood
likely to have comorbid or co-occurring psychological disor- trauma exposure on their ability to appraise danger accurately
ders or symptoms (Putnam et al., 2013). Of the 149 mothers (Pynoos, Steinberg, & Piacentini, 1999). It is also possible
in this sample who experienced at least one subtype of that mothers who are able to retain memories of feeling
childhood maltreatment, 75.8% experienced more than one safe, protected, and loved within an overall context of adver-
subtype. There may be something about angel memories sity have less psychological need for rigidly segregated men-
that is particularly helpful in protecting parents who have tal systems (Bowlby, 1980). Mothers with greater access to
experienced extensive or complex childhood trauma, such more positive and elaborated angel memories may also retain
as multiple maltreatment subtypes, against outcomes pertain- a greater ability to protect their children from their own angry
ing to traumatic stress or psychiatric comorbidities. This impulses, to detect potential danger from the environment,
tentative explanation warrants replication in other samples. and to shield their children from these threats.
Another possibility for the lack of significant buffering ef- This general finding that more positive and elaborated an-
fects of angel memories for mothers’ depression symptoms gel memories deter the transmission of trauma from mothers
may be that depression symptoms may, at least in this sample, to children is aligned with other studies that have also docu-
be characterized by a higher degree of equifinality (i.e., stem- mented parental characteristics as intergenerational protective
ming from multiple, transactional origins; Cicchetti & Ro- factors for children’s well-being. For example, maternal
gosch, 1996; Harrington, Rutter, & Fombonne, 1996). In reflective functioning (RF), defined as parents’ capacity to
the present sample, it is possible that other factors, such as reflect on their own and their children’s mental states, was
contemporaneous contextual stress or negative relationships, found to mediate associations between adult attachment
184 A. J. Narayan et al.

representations and parental attachment to infants (Slade, protection, or reassurance. Scores of 3 or below, characterized
Grienenberger, Bernbach, Levy, & Locker, 2005). While as- by vague or transient responses where the affect was not di-
sociations between angel memories and adult attachment or rectly or specifically connected to specific moments that
RF have not yet been established, it is likely that these protec- were enduring, were not effective buffers against maladapta-
tive constructs are interrelated. Parents who have more secure tion. The presence of these thresholds echoes Selma Frai-
and positive memories of their childhood experiences with berg’s observations that traumatized individuals who are
caregivers may be more likely to reflect on their mental states able to connect affect to experience are less likely to perpe-
and relationships with their children, evincing stronger trate trauma across generations (Fraiberg et al., 1975). In ad-
RF, and to respond to children with higher levels of sensitiv- dition, the findings help expand her insights by showing that
ity and protection. it may be just as important to connect positive affect to pos-
itive memories as it is to have integrated emotional narratives
of traumatic experiences.
Sex-specific protective effects of angel memories
The thresholds for protective effects observed in this study
The buffering effect of mothers’ angel memories for lower are different from those in the pilot sample. There, angel
offspring trauma exposure was significant for daughters but memories receiving scores 2 were effective buffers against
not sons. Although the specific mechanisms underlying this the association between maternal maltreatment and PTSD
protective process are still unknown, it is possible that symptoms, suggesting that even transient, isolated, or fleeting
mothers may view their daughters as more vulnerable and positive childhood memories were protective factors for
act more protectively toward them, partly because of their mothers reared in foster care (Narayan et al., 2017). Taken to-
gender-based identification and partly due to prevailing so- gether, findings across the two studies indicate that thresholds
cial perceptions of gender differences. It is also possible for angel memories to be effective protective factors may vary
that mothers may be more likely to engage in negative attribu- depending on the characteristics of the sample, the type of
tions of their sons based on their violent or otherwise acrimo- childhood adversity, and the outcomes of interest.
nious experiences with the child’s father, a pattern that has There could be several explanations for these shifting
been clinically associated with less protective behavior to- thresholds. For adults whose childhood traumatic experiences
ward sons (Lieberman, 1997, 1999). Although the mecha- significantly outweighed the positive and loving childhood
nisms at work cannot be gleaned from the present data, pre- moments, lower scores on angel memories may carry more
vious research also suggests that “matched trauma” (i.e., a weight as protective factors. The pilot sample consisted of a
child’s experience of the same trauma type that the parent ex- particularly high-risk group of mothers, all of whom had
perienced as a child) has unique effects on the parents’ per- been in childhood foster care and many of whom reported ex-
ceptions and views of the child’s well-being compared to cir- periencing very severe and chronic childhood maltreatment.
cumstances where the parent’s and child’s trauma are not It is possible that for those women, having some trace of a
“matched” (Cohodes, Hagan, Narayan, & Lieberman, loving memory, even if isolated or fleeting, served as a ray
2015). Although tentative and warranting replication, the of hope that sustained their self-worth and offered some
finding that daughters are more strongly buffered by angel protection from trauma-related impairment. The issue of
memories suggest that there may be sex-specific intergenera- how angel memories differ in their protective strength across
tional pathways of resilience. The mechanisms underlying samples with varying levels of adversity is a compelling
them are a viable area to pursue in future studies. area to pursue. An additional consideration apparent in both
studies is that very high levels of unresolved childhood
trauma might block access to any angel memories, resulting
Thresholds for protective effects of angel memories
in scores of 0 on angel memories, which were present in a
The interaction analyses provided information on cutpoints of small group of mothers in each sample. This raises the possi-
angel memories scores that buffered the effects of mothers’ bility that the status of childhood trauma as resolved versus
childhood maltreatment on their own and their children’s unresolved might also be an important consideration for fu-
well-being. While angel memories receiving scores of either ture studies.
4 or 5 were effective protective factors that buffered the ef- In addition, it is important to note that although protective
fects of maltreatment on maternal PTSD symptoms and chil- effects of angel memories for several trauma-related out-
dren’s trauma exposure, angel memories needed to receive comes were established, all protective effects in the current
the highest possible score of 5 to be effective buffers against study were modest. The interaction of maternal maltreatment
maternal comorbid psychopathology. These observations and angel memories added only a small although significant
suggest that angel memories may need to be particularly amount of variance to models predicting maternal PTSD
strong to buffer the effects of childhood maltreatment on psy- symptoms, comorbid psychopathology, and children’s
chological comorbidities. Specifically in this sample, angel trauma exposure. While statistical effects are small, it may
memories needed to be rich in emotional content, reflected be practically meaningful to consider the above thresholds
by a memory with vivid sensory details, and/or a memory when evaluating whether effects of angel memories are clini-
that connected specific moments to feelings of love, comfort, cally significant buffers against outcomes of interest.
Intergenerational protective effects of angels in the nursery 185

It is also noteworthy that angel memories did not seem to to children and that may have also been related to stronger rec-
be directly compromised by levels of psychopathology symp- ollection of angel memories. For instance, studies have found
toms. The correlations between maternal PTSD, depression, that having a supportive partner or coparent and making
and comorbid psychopathology with angel memories were meaning of trauma through therapeutic support may deter
all nonsignificant, indicating that higher levels of psychopa- the intergenerational transmission of maltreatment to children
thology do not seem to directly impair the recollection of an- (Egeland & Erickson, 2004; Egeland et al., 1988). Although
gel memories. Furthermore, the association between mothers’ the present mothers were about to enter therapy, we did not
childhood maltreatment and their angel memories was also assess whether mothers had previously participated in ther-
nonsignificant, similar to the pilot sample (Narayan et al., apy or if they had supportive partners. Several additional un-
2017). This suggests that level of overall maltreatment also measured constructs, such as reflective functioning, other
does not seem to directly compromise angel memory quality. cognitive capacities, and general emotional awareness, might
However, higher levels of maternal stress, a covariate to have also helped account for the buffering effects. Future
account for potential maternal reporting biases, were signifi- studies should investigate the additive contribution of these
cantly but modestly associated with lower angel memories, capacities in conjunction with angel memories in mitigating
suggesting that general stress levels may slightly hinder the the intergenerational transmission of childhood trauma.
recall of higher quality angel memories. Future studies should Future studies also need to examine more nuanced dimen-
directly examine associations between mothers’ angel mem- sions of angel memories, such as whether their protective ef-
ories and other distal and proximal factors, such as objective fects are weakened if the source of the angel memory is the
levels of positive early caregiving and other factors that may same individual who perpetrated maltreatment. In this study,
be associated with angel memories, including adult attach- we did not have enough information about who perpetrated
ment representations, reflective functioning, or presence of maltreatment during the mothers’ childhoods. We are cur-
unresolved trauma. rently working to differentiate, in subsequent samples, pro-
tective effects of angel memories that are characterized by
protection in the context of harm (e.g., when one caregiver
Strengths, limitations, and future directions
shields or protects the respondent from another caregiver’s
This is the first study to show that angel memories provide pro- abuse), versus rupture in the protective shield (e.g., when
tective effects across numerous maternal and child outcomes in the source of the angel memory is the same person who
a sample of ethnically, culturally, and linguistically diverse also perpetrated harm). It is likely that these two circumstan-
families. A key strength of the study is that it illustrates the util- ces would result in different forms of psychological adapta-
ity of an instrument that is brief to administer and code, yet as- tion. Psychological maladaptation may be especially likely
sesses a replicable protective factor against lifespan and interge- if the respondent is still haunted by a caregiver’s contradic-
nerational transmission of trauma. Additional strengths include tory behavior. Alternatively, better psychological adaptation
a relatively large sample, the ability to control for mothers’ would likely follow from angel memories that are character-
stress levels given that they were the primary informants on ized by recollections of positive, loving moments that
all variables, and replicability of findings in this sample despite do not contain elements of maltreatment or references to
utilization of different instruments than the pilot sample. traumatic experiences at all.
There were also several limitations. The replicability of
findings involves primarily Latina mothers, a quarter of
Conclusions
whom were monolingual Spanish speaking. While this is
an important first step in documenting the cross-cultural util- The present study provides support for the concept that angels
ity of angel memories as protective factors when compared in the nursery are effective protective factors that buffer
with the pilot sample of primarily African American mothers, against the intergenerational transmission of trauma in a
future studies should examine angel memories in families of group of predominantly low-income, ethnically diverse
additional racial and ethnic backgrounds. In addition, this mothers and young children. These protective effects have
sample was clinically referred and characterized by high now been replicated across two independent samples with
levels of contemporaneous maternal stress (the sample variance in methods; ethnic, cultural, and linguistic diversity;
mean on maternal stress corresponded to the 85th percentile and types of childhood adversity and psychopathology. An-
of the Parenting Stress Index, Short Form, Version 3, accord- gels in the nursery are particularly advantageous because
ing to developers). While it is likely important to utilize they can be identified and coded briefly with minimal training
samples with high levels of, and high variation in, childhood and low burden to participants and researchers. They offer
adversity and ongoing stress in order to examine protective parents who have experienced trauma an opportunity to
factors that deter maladaptation, it is unknown whether angel reflect on moments of childhood love, happiness, and protec-
memories would exert similar protective effects in lower tion. Mothers administered the Angels Interview often
stress, non-clinically referred samples. comment afterward, “I’ve never really thought about a posi-
This study did not take into account additional factors that tive memory,” “No one has ever asked me about this before,”
may have buffered the transmission of trauma from mothers and “That was very helpful, thank you.” Future research is
186 A. J. Narayan et al.

needed to continue to investigate angels in the nursery as pro- daptation and whether angel memories can be systematically
tective factors that buffer against the effects of punitive or nurtured and strengthened with evidence-based, trauma-
abusive caregiving on lifespan and intergenerational mala- informed interventions.

References
Abidin, R. R. (1990). Parenting Stress Index (Short Form). Charlottesville, Hayes, A. F., & Matthes, J. (2009). Computational procedures for probing inter-
VA: Pediatric Psychology Press. actions in OLS and logistic regression: SPSS and SAS implementations. Be-
Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Manual for the Beck De- havior Research Methods, 41, 924–936. doi:10.3758=BRM.41.3.924
pression Inventory—II. San Antonio, TX: Psychological Corporation. Holt, S., Buckley, H., & Whelan, S. (2008). The impact of exposure to do-
Beck, A. T., Steer, R. A., & Carbin, M. G. (1988). Psychometric properties of mestic violence on children and young people: A review of the literature.
the Beck Depression Inventory: Twenty-five years of evaluation. Clinical Child Abuse & Neglect, 32, 797–810. doi:10.1016/j.chiabu.2008.02.004
Psychology Review, 8, 77–100. doi:10.1016/0272-7358(88)90050-5 Lieberman, A. F. (1997). Toddlers’ internalization of maternal attributions as
Bernstein, D. P., Stein, J. A., Newcomb, M. D., Walker, E., Pogge, D., Ah- a factor in quality of attachment. In L. Atkinson & K. J. Zucker (Eds.),
luvalia, T., . . . Zule, W. (2003). Development and validation of a brief Attachment and psychopathology (pp. 277–291). New York: Guilford
screening version of the Childhood Trauma Questionnaire. Child Abuse Press.
& Neglect, 27, 169–190. doi:10.1016/S0145-2134(02)00541-0 Lieberman, A. F. (1999). Negative maternal attributions: Effects on toddlers’
Bowlby, J. (1980). Attachment and loss: Vol. 3. Loss: Sadness and depres- sense of self. Psychoanalytic Inquiry, 19, 737–756. doi:10.1080/
sion. London: Hogarth Press and the Institute of Psycho-Analysis. 07351699909534274
Bowlby, J. (1988). A secure base: Parent-child attachment and healthy hu- Lieberman, A. F., Padrón, E., Van Horn, P., & Harris, W. W. (2005). Angels
man development. New York: Basic Books. in the nursery: The intergenerational transmission of benevolent parental
Briere, J., Elliott, D. M., Harris, K., & Cottman, A. (1995). Trauma Symptom influences. Infant Mental Health Journal, 26, 504–520. doi:10.1002/
Inventory: Psychometrics and association with childhood and adult vic- imhj.20071
timization in clinical samples. Journal of Interpersonal Violence, 10, Lieberman, A. F., & Van Horn, P. (2008). Psychotherapy with infants and
387–401. doi:10.1177/088626095010004001 young children. New York: Guilford Press.
Cicchetti, D., & Rogosch, F. A. (1996). Equifinality and multifinality in de- Main, M., & Goldwyn, R. (1984a). Adult attachment scoring and classifica-
velopmental psychopathology. Development and Psychopathology, 8, tion systems (2nd ed.). Unpublished manuscript, University of California,
597–600. Berkeley.
Cohodes, E., Hagan, M., Narayan, A. J., & Lieberman, A. F. (2015). Matched Main, M., & Goldwyn, R. (1984b). Predicting rejection of her infant from
trauma: The role of parents’ and children’s matched experiences of child- mothers’ representation of her own experience: Implications for the
hood trauma in parents’ report of children’s trauma-related symptomatol- abused-abusing intergenerational cycle. Child Abuse & Neglect, 8, 203–217.
ogy. Journal of Trauma and Dissociation, 17, 81–96. doi:10.1080/ McCloskey, L. A., & Bailey, J. A. (2000). The intergenerational transmission
15299732.2015.1058878 of risk for child sexual abuse. Journal of Interpersonal Violence, 15,
Cook, D. R., & Weisberg, S. (1982). Residuals and influence in regression. 1019–1035. doi:10.1177/088626000015010001
London: Chapman & Hall. Narayan, A. J., Ghosh Ippen, C., Harris, W. W., & Lieberman, A. F. (2017).
Davidson, J. R. T., Book, S. W., Colket, J. T., Tupler, L. A., Roth, S., David, Assessing angels in the nursery: A pilot study of childhood memories of
D., . . . Feldman, M. (1997). Assessment of a new self-rating scale for benevolent caregiving as protective influences. Infant Mental Health
post-traumatic stress disorder. Psychological Medicine, 27, 153–160. Journal, 38, 461–474. doi:10.1002/imhj.21653
doi:10.1017/S0033291796004229 Narayan, A. J., Oliver Bucio, G., Rivera, L. M., & Lieberman, A. F. (2016).
Egeland, B., & Erickson, M. F. (2004). Lessons from STEEP: Linking theory, Making sense of the past creates space for the baby: Perinatal child-parent
research, and practice on the well-being of infants and parents. In A. Sa- psychotherapy for pregnant women with childhood trauma. Zero to Three
meroff, S. McDonough, & K. Rosenblum (Eds.), Treating parent–infant Journal, 36, 22–28.
relationship problems (pp. 213–242). New York: Guilford Press. Narayan, A. J., Rivera, L. M., Bernstein, R. E., Harris, W. W., & Lieberman,
Egeland, B., Jacobvitz, D., & Sroufe, L. A. (1988). Breaking the cycle of A. F. (2018). Positive childhood experiences predict less psychopathol-
abuse. Child Development, 59, 1080–1088. doi:10.2307/1130274 ogy and stress in pregnant women with childhood adversity: A pilot study
Ertem, I. O., Leventhal, J. M., & Dobbs, S. (2000). Intergenerational conti- of the benevolent childhood experiences (BCEs) scale. Child Abuse &
nuity of child physical abuse: How good is the evidence? Lancet, 356, Neglect. Advance online publication. doi:10.1016/j.chiabu.2017.09.022
814–819. doi:10.1016/S0140-6736(00)02656-8 Putnam, K. T., Harris, W. W., & Putnam, F. W. (2013). Synergistic childhood
Fonagy, P., Gergely, G., & Target, M. (2008). Psychoanalytic constructs and adversities and complex adult psychopathology. Journal of Traumatic
attachment theory and research. In J. Cassidy & P. Shaver (Eds.), Hand- Stress, 26, 435–442. doi:10.1002/jts.21833
book of attachment (2nd ed., pp. 783–810). New York: Guilford Press. Pynoos, R. S., Steinberg, A. M., & Piacentini, J. C. (1999). Developmental
Fraiberg, S. (1980). Clinical studies in infant mental health: The first year of psychopathology of childhood traumatic stress and implications for asso-
life. New York: Basic Books. ciated anxiety disorders. Biological Psychiatry, 46, 1542–1554.
Fraiberg, S., Adelson, E., & Shapiro, V. (1975). Ghosts in the nursery: A psy- doi:10.1016/S0006-3223(99)00262-0
choanalytic approach to the problem of impaired infant-mother relation- Raby, K. L., Steele, R. D., Carlson, E. A., & Sroufe, L. A. (2015). Continu-
ships. Journal of the American Academy of Child & Adolescent Psychia- ities and changes in infant attachment patterns across two generations. At-
try, 14, 387–421. doi:10.1016/S0002-7138(09)61442-4 tachment and Human Development, 17, 414–428. doi:10.1080/
Geeraert, L., van den Noortgate, W., Grietens, H., & Onghena, P. (2004). The 14616734.2015.1067824
effects of early prevention programs for families with young children at Rawlings, J. O. (1988). Applied regression analysis: A research tool. Pacific
risk for physical child abuse and neglect: A meta-analysis. Child Mal- Grove, CA: Wadsworth & Brooks-Cole.
treatment, 9, 277–291. doi:10.1177/1077559504264265 Roisman, G. I., Padrón, E., Sroufe, L. A., & Egeland, B. (2002). Earned–se-
Ghosh Ippen, C., Ford, J., Racusin, R., Acker, M., Bosquet, M., Rogers, K., cure attachment status in retrospect and prospect. Child Development, 73,
. . . Edwards, J. (2002). Traumatic Events Screening Inventory—Parent 1204–1219. doi:10.1111/1467-8624.00467
Report Revised. Hanover, NH: US Department of Veteran’s Affairs, Na- Rubin, D. B. (1987). Multiple imputation for nonresponse in surveys.
tional Center for PTSD. New York: Wiley.
Ghosh Ippen, G., Narayan, A. J., Van Horn, P., & Lieberman, A. F. (2015). Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state of the
Manual for coding angels and ghosts in the nursery. University of Cali- art. Psychological Methods, 7, 147–177. doi:10.1177/0049124102030003008
fornia, San Francisco, Child Trauma Research Program. Shaffer, A., & Sroufe, L. A. (2005). The developmental and adaptational
Harrington, R., Rutter, M., & Fombonne, E. (1996). Developmental pathways in implications of generational boundary dissolution: Findings from a
depression: Multiple meanings, antecedents, and endpoints. Development prospective, longitudinal study. Journal of Emotional Abuse, 5, 67–84.
and Psychopathology, 8, 601–616. doi:10.1017/S095457940000732X doi:10.1300/J135v05n02_04
Intergenerational protective effects of angels in the nursery 187

Slade, A., Grienenberger, J., Bernbach, E., Levy, D., & Locker, A. (2005). Van Horn, P., Lieberman, A. F., & Harris, W. W. (2008). The Angels in the
Maternal reflective functioning, attachment, and the transmission gap: Nursery Interview. San Francisco, CA: University of California,
A preliminary study. Attachment and Human Development, 7, 283– San Francisco, Department of Psychiatry.
298. doi:10.1080/14616730500245880 Widom, C. S., & Wilson, H. W. (2015). Intergenerational transmission of
Spitzer, R. L., Williams, J. B. W., Gibbon, M., & First, M. B. (1990). violence. In J. Lindert & I. Levav (Eds.), Violence and mental health
Structured Clinical Interview for DSM-III-R (SCID). New York: (pp. 27–45). Dordrecht, Netherlands: Springer.
New York State Psychiatric Institute, Biometrics Research Wolfe, J. W., Kimerling, R., Brown, P. J., Chrestman, K. R., & Levin, K.
Department. (1996). Psychometric review of the Life Stressor Checklist—Revised.
Sroufe, L. A., Egeland, B., Carlson, E. A., & Collins, W. A. (2005). The In B. H. Stamm (Ed.), Measurement of stress, trauma, and adaptation
development of the person. New York: Guilford Press. (pp. 198–201). Lutherville, MD: Sidran Press.