J Abnorm Child Psychol

DOI 10.1007/s10802-014-9937-y

Self-Compassion in Recovery Following Potentially Traumatic

Stress: Longitudinal Study of At-Risk Youth
Mordechai Zeller & Kim Yuval & Yaara Nitzan-Assayag &
Amit Bernstein

# Springer Science+Business Media New York 2014

Abstract Despite promising theory, empirical study of the 2000; Brewin and Holmes 2003; Elwood et al. 2009). Such
putative protective properties of self-compassion (SC) with knowledge is essential for guiding the development of trauma-
respect to resilience to and recovery from traumatic stress is related prevention and early intervention (Feldner et al. 2007).
limited. The present study tested the theorized protective A range of factors have been theorized as relevant to trauma
role(s) of SC with respect to trauma-related psychopathology resilience and recovery, yet have received relatively limited
over time among an at-risk sample of adolescents (N=64, direct empirical study. One such promising construct is self-
26 % females, M(SD) age =17.5(1.07) years-old, range age = compassion (SC). Accordingly, the present study explores the
15–19; grades 9–12) directly exposed to a potentially traumat- role(s) SC in recovery following traumatic stress.
ic stressful event – the Mount Carmel Forest Fire Disaster. The
longitudinal design involved three assessment time-points – Compassion & Self-Compassion Growing attention to com-
within 30-days of the potentially traumatic event (T1) and passion in the clinical psychological science literature is
then at 3- (T2) and 6-months (T3) follow-up intervals. linked to the field’s broader adoption of principles and tech-
Consistent with prediction, multi-level modeling of mediation niques central to Buddhist traditions (Hayes et al. 2011;
documented the prospective protective function(s) of SC, Hofmann et al. 2011). Compassion is one of four
above and beyond dispositional mindfulness, with respect to Brahmaviharas also known as the four immeasurables – com-
posttraumatic stress and panic symptoms, depressive symp- passion, loving-kindness, joy, and equanimity (Kraus and
toms, and suicidality symptoms, but not well-being. The Sears 2009). In Theravada, Mahayana, and Tibetan Buddhist
findings are discussed, theoretically, with respect to SC as a traditions, compassion is broadly conceptualized as the desire
malleable protective factor for trauma-related psychopatholo- to prevent suffering from happening to others and one’s self
gy outcomes; and, clinically, with respect to SC as a target for (Davidson and Harrington 2001; Tirch 2010). In contempo-
future trauma-related selective-prevention and -early interven- rary psychological literature, a number of theorists have pro-
tion research. posed alternative, albeit related, conceptual models of com-
passion and SC broadly consistent with the Buddhist concep-
Keywords Adolescents . Compassion . Developmental tualization (Gilbert and Procter 2006; Neff 2003a; Ozawa-de
psychopathology . Longitudinal . Protective factor . Silva et al. 2012). For example, Neff (2003a) conceptualized
Self-compassion . Stress . Trauma . Risk factor . Youth SC as feelings of care and kindness toward oneself through
taking an understanding, non-judgmental attitude towards
one’s perceived inadequacies, recognizing that one’s personal
We have much to learn about malleable causal risk and pro- experience is connected to the greater collective human expe-
tective factors underlying the development of trauma-related rience and by a willingness to be open to one’s own suffering
psychopathology including posttraumatic stress, depression, without avoiding it (see also Neff 2003b).
panic, and related disorders (Bomyea et al. 2012; Brewin et al.
Self-Compassion and Recovery from Traumatic Stress: Con-
M. Zeller : K. Yuval : Y. Nitzan-Assayag : A. Bernstein (*)
ceptual Rationale We theorize that the more compassionately
Department of Psychology, University of Haifa, Mount Carmel,
Haifa 31905, Israel a person relates towards her/his self following exposure to
e-mail: [email protected] traumatic stress, the more likely that person may be to: (a)
 J Abnorm Child Psychol

recognize the need to care for oneself and thereby engage in prospective protective role(s) of SC following exposure to
self-care behaviors or benefit from social support (e.g., Allen major life or potentially traumatic stress on trauma-related
and Leary 2010; Neff 2003b); (b) permit oneself to experience outcomes.
rather than avoid painful thoughts, emotions and physical
sensations often experienced post-trauma and thereby facili- Experimental Laboratory Study A number of controlled lab-
tate emotional processing of the event (e.g., Thompson and oratory studies have begun to test the effects of exper-
Waltz 2008); (c) engage in less ruminative self-blame or imentally manipulating compassion and SC (i.e., via
related self-criticism (e.g., Gilbert and Procter 2006; Leary Loving Kindness Meditation or Compassion Meditation) on
et al. 2007; Raes 2010); and (d) experience the memory of the a variety of affective, social and neural outcomes (Hutcherson
event from a more decentered or self-distanced perspective – et al. 2008; Lutz et al. 2004, 2008). With respect to stress and
as an emotionally painful experience rather than as an experi- trauma, Leary and colleagues (2007) instructed partici-
ence that defines or changes oneself per se (e.g., Kross and pants to recall and describe a negative autobiographical
Ayduk 2011; Neff 2003b). Consistent with such theorizing, memory of a past event experienced as a failure.
there is a growing body of empirical research documenting the Relative to an active self-esteem condition and two
salutary role of SC for stress-related disorders such as mood control conditions, participants randomized to the SC
and anxiety disorders (Neff 2003a; Neff et al. 2007; Raes condition reported lower levels of negative affect and
2011), as well as preliminary evidence of its protective role were more likely to acknowledge their role in the event
with respect to trauma-related psychopathology more specif- (Leary et al. 2007). To the best of our knowledge, no
ically (Thompson and Waltz 2008). We focus on the latter other experimental investigation has directly tested the
below. protective role(s) of dispositional SC or experimentally
manipulated levels of SC on responding to a novel
Cross-Sectional Study In addition a to a number of cross- stressor in the lab, such as on emotional recovery or
sectional studies linking SC to depression, anxiety and related avoidance behavior.
symptoms (Neff 2003a; Neff et al. 2007; Neff et al. 2008; Neff
and McGehee 2010; Raes 2010; Van Dam et al. 2011; see Intervention & Prevention Study A number of intervention
MacBeth and Gumley 2012 for meta-analytic review), a num- studies have begun to test the salutary effects of compassion
ber of studies have begun to evaluate the association between and SC on mood and anxiety-related outcomes (i.e.,
SC and trauma-related psychopathology specifically. Fredrickson et al. 2008; Gilbert and Irons 2004, Gilbert and
Thompson and Waltz (2008) found that SC was related to Procter 2006; Neff et al. 2007; Shapira and Mongrain 2010),
reduced levels of PTSD avoidance symptom severity specif- as well as on responding to stress or trauma-related stress
ically among a non-clinical trauma-exposed sample more specifically. Reddy and Colleagues (2012) found, in an
(Thompson and Waltz 2008). Tanaka et al. (2011) found that uncontrolled test, that a 6-week Cognitive Based Compassion
higher levels of childhood emotional abuse, emotional neglect Training program (CBCT), delivered to at-risk adolescents in
and physical abuse were associated with lower levels of SC; foster care, led to elevations in hopefulness and reduced levels
and that adolescents with lower SC levels were more likely to of generalized anxiety as a function of the frequency of CBCT
suffer from psychological distress leading to problems with practice. In a randomized control design, Pace and colleagues
substance abuse and in some instances attempted suicide (2009) found that relative to a health discussion control con-
among a sample of at-risk youth receiving child protective dition, compassion meditation led to improved immune (plas-
services. Relatedly, Vettese et al. (2011) found that SC con- ma concentrations of interleukin IL-6 related to down-stream
currently mediated the association between childhood inflammation and disease), neuroendocrinological (cortisal),
maltreatment and later adolescent emotional dysregulation in and subjective distress. However, at 2–4 weeks post-interven-
a sample of youth seeking treatment for substance abuse tion, no group differences were observed with respect to
above and beyond other risk factors including childhood responding to a laboratory-based psychosocial stressor (Trier
maltreatment history, current psychological distress and Social Stress Test). Furthermore, greater levels of CM practice
addiction severity. among the CM condition were related to reduced IL-6 and
distress scores, but not cortisol levels, in response to the
Longitudinal Study To the best of our knowledge, only one stressor. Finally, in a randomized control study, participants
longitudinal study has evaluated relations between SC, psy- seeking therapy following a traumatic event were randomized
chopathology, and related outcomes. Raes (2011) found that to CBT or CBT-plus-CMT (Beaumont et al. (2012). Both
among a large non-clinical university student sample, greater intervention conditions demonstrated significant reductions
levels of SC at one time-point predicted reduced levels of in depression, anxiety, and other posttraumatic stress symp-
depression symptoms at a 5-month follow-up (Raes 2011). toms; though the CBT-CMT group also demonstrated eleva-
To the best of our knowledge, no study has yet to test the tions in SC. To the best of our knowledge, no prevention study
J Abnorm Child Psychol

targeting SC has been conducted to-date; nor has an early groups within the institution (e.g., grades) so as to methodo-
intervention study tested the preventive role of SC for resil- logically maximize the independence of observations between
ience or recovery from trauma or related stress exposure. sampled youth appropriate for a 2-level mixed multi-level
model design (i.e., level 1 = repeated measurements, level 2
Gaps in Extant Research on Self-Compassion and Trauma = participants).1 The youth village is not a mental health
Recovery Research has yet to test the role(s) of SC preceding treatment facility but a residential educational facility.
or immediately following traumatic stress exposure for psy- 92.1 % of the sample were Ethiopian Jews, 6.3 % Jews from
chopathology risk. Developmentally-oriented, prospective the Former Soviet Union, and one participant (1.6 %) was not
study is particularly needed – to test the putative protective an immigrant. None were forcibly displaced persons, refugees
role(s) of SC for the development and maintenance of stress- or asylum seekers. All spoke and read Hebrew fluently. In
and trauma-related psychopathology outcomes. Furthermore, terms of socio-economic status, 61 % of the sample reported
in light of the developmental emergence and long-term impli- “much below average” and an additional 22 % “below aver-
cations of stress-related disorders among youth, and at-risk age”. The youth village reports that approximately 50 % of
youth in particular (Costello et al. 2006; Grant et al. 2004), a their students originate from homes struggling with various
developmental psychopathology perspective on SC and trau- chronic stressors including poverty, violence, and substance
ma recovery may be particularly informative. Theoretically, abuse; and approximately 20 % of the students are orphans.
by reducing the effects of (life-long) retrospective censored Consistent with conceptualization of the stressor as potentially
observations (e.g., past traumatic stress exposure, psychiatric traumatic, at T1 (within 30-days of the fire) 88 % participants
history, etc.) among adults, that may bias prospective study of reported that they feared that their life and/or the lives of those
SC and trauma-related adjustment, study among youth may be close to them were threatened (e.g., friends, relatives), and/or
particularly informative. Clinically, prospective study among felt helpless and/or horror over the course of the fire (see
youth at elevated risk of poor adjustment post-trauma (i.e., at- below for details).
risk youth) may be important for identifying candidate mal- The study received human subjects research ethics approv-
leable risk/protective factors, such as SC, to target in novel al through a University of Haifa IRB committee. Informed
trauma-related selective-prevention or -early intervention written consent for participation was provided, first, by the
programs. institutional director and legal guardian of the youth; and
second, by each adolescent participant. Participants received
Specific Aims Accordingly, we tested the prospective associ- up to 120NIS ($30) in exchange for participation in a total of
ation between SC and resilience to and recovery from a recent three 1-hour assessment sessions.
(past 30-day) potentially traumatic stressor at three time-
points over a 6-month time-period, among a sample of Procedure
at-risk adolescents. Specifically, we tested the extent to
which SC prospectively mediated the effect of time, Participants were recruited immediately following the Carmel
following exposure to the traumatic stressor, on Fire Disaster—a large, week-long, forest fire (December,
transdiagnostic trauma-related psychopathology symp- 2010) that led to emergency evacuation and destruction of
toms, including posttraumatic stress, depression, panic, local communities, including injuries and deaths (Bronner
and suicidality symptoms, as well as well-being. We 2010). The youth village was evacuated and about 40 % of
also tested the incremental predictive validity of SC the structures were destroyed by the fire, displacing the youth
above and beyond the effect of dispositional mindful during the week-long fire and in subsequent weeks.
attention and awareness, an alternative and theoretically Recruitment was initiated only once the fire no longer acutely
related protective factor (Tirch 2010; Van Dam et al. threatened local communities including the youth village.
2011). Study inclusion criteria included: (a) a minimum Hebrew
reading at sixth grade-level; (b) proximal exposure to the
recent (past 30-day) potentially traumatic stressor. The first
Method
1
Students belonged to multiple within-grade class units, such that only 4
possible level-3 groups (i.e., grades) may be tested. There were therefore
Participants
an insufficient number of potential level-3 units to test the relative
improved fit of a 3- relative to 2-level models. An insufficient number
Data were collected from a convenience sample of 64 high- of groups at level-3 relative to parameters to be estimated results in
school students (26.6 % female; M(SD)age =17.5(1.07) years- problems with model convergence, unreliable level-3 parameter esti-
mates, and underpowered level 3 analyses (Singer and Willett 2003;
old, rangeage =15–19; grades 9–12) living in an educational
Tasca et al. 2009). Accordingly, sampling of youth was designed a priori
residential youth village in northern Israel. Participants were to maximize independence of observations for a 2- rather than 3-level
sampled across the entire institution, rather than from nested model.
 J Abnorm Child Psychol

assessment session was held within 4-weeks of (following) reliability and validity (Neff 2003a, b). As in extant study of
the fire (T1), and then ~3-months post- (T2), and ~6-months SC, we utilized the SCS-total score rather than separate di-
post-fire (T3). At T2, 88 % of participants who participated at mensions or sub-scales because it best reflects the SC con-
T1 were retained, and at T3, 70 % of participants who partic- structs (Neff 2003a) and to reduce the number of analyses.
ipated at T1 and 80 % of participants who participated at T2 Internal consistency in these data were acceptable (Cronbach’s
were retained. The retention rate would have likely been α=0.64), but lower than some earlier reported levels (Neff
higher but a number of participants graduated and drafted to 2003a).
military service between T2 and T3. Moreover, observed
prospective retention rates may be considered high in light The Inventory of Depression and Anxiety Symptoms (IDAS;
of the nature of the high-risk youth population and fully Watson et al. 2007) The IDAS is a factor-analytically derived,
sufficient to conduct the proposed mixed multi-level modeling multidimensional inventory that uses a 5-point Likert-type
analyses reliably (Singer and Willett 2003). scale (1 = not at all to 5 = Extremely) to assess current
Participants were invited to the school computer laboratory symptoms. In the proposed study, we focus a priori on symp-
where they completed the assessment (~n=10 per group). toms of psychopathology related to traumatic stress, including
Participants were seated at a sufficient distance (not at adja- general depression, suicidality, panic, posttraumatic stress and
cent computer stations) from one another to ensure their well-being sub-scales, experienced in the past 2-weeks. The
privacy. School staff were not present during the assessment IDAS has demonstrated strong internal consistency, test-retest
session, and full confidentiality of participant’s reports was reliability, and good convergent and discriminant validity with
guaranteed to youth participants to facilitate sound self-report. respect to formal diagnostic and self-report symptom mea-
Participants were told that their responses would not be shared sures in multiple populations (Watson et al. 2007). Internal
in any way with the institutional staff, unless a participant consistency of the IDAS sub-scale scores were good to excel-
reported that she/he was in immediate physical danger (e.g., lent (Cronbach’s α’s=0.79 to 0.93), consistent with past work
serious suicidal ideation). Members of the research team were (Watson et al. 2007).
present during all assessment sessions.
Mindful Attention Awareness Scale (MAAS; Brown and Ryan
Measures 2003) The MAAS is a 15-item questionnaire in which respon-
dents indicate, on a 6-point Likert-type scale (1 = almost
Translation and Back-Translation Process The measures always to 6 = almost never), their level of dispositional aware-
were translated from English to Hebrew by laboratory staff ness and attention to present events and experiences (Brown
fluent in Hebrew and English. The scales were then back- and Ryan 2003). Sample MAAS items include “I rush through
translated by a separate party using structured guidelines activities without being really attentive to them” and “I find it
(Geisinger 1994). difficult to stay focused on what’s happening in the present.”
The MAAS has demonstrated good internal consistency
Descriptive Demographic Measures Participants provided across a range of samples (α=0.80–0.87), and strong test-
demographic and related personal background information retest reliability data over a 1-month time period (r=0.81;
(religion, race/ethnicity, age, gender). Brown and Ryan 2003). Internal consistency of the MAAS
total score was excellent (Cronbach’s α=0.88), consistent
Self Compassion Scale (SCS; Neff 2003a) SCS is a 26-item with past work (Brown and Ryan 2003).
self report questionnaire consisting of a total score composed
of six dimensions of SC including Self-kindness (“When I’m Carmel Trauma Questionnaire (CTQ) The CTQ is an 8-item
going through a very hard time, I give myself the caring and self-report questionnaire, developed in our laboratory for the
tenderness I need”), Self-judgment (“When times are really purpose of this study to assess relevant information related to
difficult, I tend to be tough on myself”), Common humanity the Carmel fire disaster specifically, such as proximity to the
(“I try to see my failings as part of the human condition”), fire, injury to the participant or relatives, property damage and
Isolation (“When I think about my inadequacies it tends to related information. In addition, participants were asked to
make me feel more separate and cut off from the rest of the answer general questions about their feelings during the fire
world”), Mindfulness (“When something upsets me, I try to (“Did you feel that your life was in danger?”). The CTQ was
keep my emotions in balance”), and Over-identification grounded in established measures of traumatic stress exposure
(“When something upsets me I get carried away with my and posttraumatic stress, including the Posttraumatic Diagnostic
feelings”). SCS items were designed to measure SC in times Scale and the Clinician Administered Posttraumatic Stress Scale
of difficulty and thus selected for the present investigation. structured interview. This measure was used only to gain de-
Items are rated on a 5-point Likert-type scale (1 = never to 5 = scriptive data regarding participants’ exposure to the potentially
almost always). The SCS has demonstrated multiple forms of traumatic stress event.
J Abnorm Child Psychol

Results and beyond the effect of SC. This model explained 15 % of the
variance in Depressive symptoms (pseudo-r2 =0.15)
Multilevel Modeling of Mediation: Data Analytic Strategy All
data analyses were carried out in SPSS. We tested multi-level Panic Symptoms Panic symptoms rose significantly over time
models of mediation in which time (T1, T2, T3) was the (β=0.31, SE=0.15, t=2.03, p<0.05). We then regressed SC
predictor, SC was the mediator, and trauma-related symptoms on time (see above). Finally, we regressed panic symptoms on
(depressive, panic, posttraumatic stress, and suicidality symp- both time and SC simultaneously. The effect of SC was
toms) and well-being were the outcomes (Kenny et al. 2003). significant, indicating that higher levels of SC at time T1
The models were lagged such that SC at time T predicted and T2 predicted lower levels of panic symptoms (i.e., less
symptom outcomes at time T + 1. The model used the diag- elevation or reductions in symptoms) at T2 and T3, respec-
onal covariance error structure matrix. The variables standard- tively (β=−2.01, SE=0.64, t=−3.16, p<0.01). The effect of
ized to facilitate interpretation of the path coefficients across time became non-significant (β=1.88, SE=1.16, t=1.62, p=
models. Finally, in the event that a null effect of time was 0.11), indicating that time did not have an additional effect on
observed, we re-ran the MLM of mediation but without the panic symptoms above and beyond the effect of SC; SC thus
lag. Because we collected data at 3 time-points, the lag re- fully mediated the effect of time on panic symptoms. This
duces the statistical power of the MLM to detect an effect of model explained 32 % of the variance in panic symptoms
time and thus a non-lagged analysis is important to rule-out a (pseudo-r2 =0.32)
null effect of time due to this potential methodological artifact
(Singer and Willett 2003). See Table 1 for descriptive statistics Posttraumatic Stress Symptoms Posttraumatic stress symp-
for all studied variables. toms rose significantly over time (β=0.31, SE=0.13, t=2.34,
p<0.05). We then regressed SC on time (see above). Finally,
Depressive Symptoms Depressive symptoms levels did not we regressed posttraumatic stress symptoms on both time and
change significantly over time (β=0.26, SE=0.16, t=1.7, p= SC simultaneously. The effect of SC was significant indicating
0.11). SC levels decreased significantly over time (β=−0.48, that higher levels of SC at time T1 and T2 predicted lower
SE=0.18, t=−2.7, p<0.01). Finally, we regressed depressive levels of posttraumatic stress symptoms (i.e., less elevation or
symptoms on both time and SC simultaneously. The effect of reductions in symptoms) at T2 and T3, respectively (β=
SC was significant, indicating that higher levels of SC at time −0.17, SE=0.08, t=−2.02, p<0.05). The effect of time be-
T1 and T2 predicted lower levels of depressive symptoms at came non-significant (β=0.21, SE=0.14, t=1.49, p=0.15),
T2 and T3, respectively (β = −0.23, SE = 0.09, t = −2.3, indicating that time did not have an additional effect on
p<0.05). The effect of time remained non-significant (β= posttraumatic stress symptoms above and beyond the effect
0.16, SE=0.16, t=0.97, p=0.35), indicating that time did not of SC; SC thus fully mediated the effect of time on posttrau-
have an additional effect on depressive symptoms above and matic stress symptoms. This model explained 14 % of the
beyond the effect of SC. The model explained 25 %2 of the variance in trauma symptoms (pseudo-r2 =0.14).
variance in Depressive symptoms (pseudo-r2 =0.25; Singer
and Willett 2003). Suicidality Symptoms Suicidality levels did not change sig-
As planned, because there was no significant change of nificantly over time (β=0.22, SE=0.16, t=1.42, p>0.05). We
depressive symptoms over time in the lagged model, we re-ran then regressed SC on time (see above). Finally, we regressed
a non-lagged MLM of mediation. Depressive symptoms suicidality on both time and SC simultaneously. The effect of
levels did not change significantly over time though a trend SC was significant, indicating that higher levels of SC at time
nearing significance was observed (β=0.17, SE=0.09, t=1.9, T1 and T2 predicted lower levels of suicidality symptoms at
p=0.06). SC levels decreased significantly over time (β= T2 and T3, respectively (β = −0.22, SE = 0.09, t = −2.38,
−0.22, SE=0.09, t=−2.4, p<0.05). Finally, we regressed de- p<0.05). The effect of time remained non-significant (β=
pressive symptoms on both time and SC simultaneously. The 0.15, SE=0.16, t=0.89, p>0.05), indicating that time did not
effect of SC was significant, indicating that higher levels of have an additional effect on suicidality symptom levels above
SC at time T1, T2 and T3 predicted lower levels of depressive and beyond the effect of SC. This model explained 15 % of the
symptoms at T1, T2 and T3, respectively (β=−0.25, SE=0.08, variance in suicidality symptoms (pseudo-r2 =0.15).
t=− 3.3, p<0.01). The effect of time remained non-significant As planned, because there was no significant change in
(β=0.12, SE=0.09, t=1.3, p=0.20), indicating that time did suicidality symptoms over time in the lagged model, we re-ran
not have an additional effect on depressive symptoms above a non-lagged model. Suicidality levels rose significantly over
time (β=0.22, SE=0.09, t=2.51, p<0.05). We then regressed
2 SC on time (see non-lagged model above). Finally, we
Pseudo-r2 =sigma-hatε2 (unconditional means model) - sigma-hatε2
(unconditional growth model) / sigma-hatε2 (unconditional means mod- regressed suicidality on both time and SC simultaneously.
el)] (Singer and Willett 2003, p. 103) The effect of SC was significant, indicating that elevation in
 J Abnorm Child Psychol

Table 1 Descriptive statistics of

studied variables over time post- Time 1 (N=64) Time 2 (N=56) Time 3 (N=45)
trauma
M (SD)
T1 < 4-weeks post-trauma; T2 = SCS - Self-compassion 84.2 (9.6) 80.1 (6.4) 81.1 (5.8)
~3-months post-trauma; T3 = ~6- MAAS - Mindfulness 67.4 (13.2) 60.6 (17.6) 64.5 (15.8)
months post-trauma. Self Compas- IDAS - Depressive symptoms 39.6 (12.5) 40.4 (12.7) 43.2 (13.8)
sion Scale (SCS; Neff 2003a); IDAS - Panic symptoms 13.2 (6.6) 12.9 (6.8) 15.8 (8.6)
Mindful Attention Awareness Scale
(MAAS; Brown and Ryan 2003); IDAS - Posttraumatic stress symptoms 7.3 (3.4) 7.0 (3.7) 8.3 (4.1)
The Inventory of Depression and IDAS - Suicidality symptoms 8.0 (3.6) 9.3 (5.3) 10.2 (5.3)
Anxiety Symptoms (IDAS; Watson IDAS - Well-being 23.5 (4.2) 23.8 (4.1) 22.5 (3.7)
et al. 2007)

SC at time T1, T2 and T3 significantly predicted reduction in 6 months following exposure to a potentially traumatic
suicidality symptoms (i.e., less elevation or reductions in stressor.
symptoms) at T1, T2 and T3, respectively (β=−0.20, SE= Consistent with prediction, multi-level modeling of medi-
0.08, t=−2.51, p<0.05). The effect of time became non- ation documented that greater levels in SC were prospectively
significant (β=0.18, SE=0.09, t=1.92, p=0.58), indicating predictive of lower levels or lesser elevations (i.e., were pro-
that time did not have an additional effect on suicidality tective) of trauma-related psychopathology symptom out-
symptoms above and beyond the effect of SC; SC thus fully comes but not well-being. Specifically, in lagged analyses,
mediated the effect of time on suicidality. This model, how- elevation in SC from T1 to T2 predicted reduction or lesser
ever, explained only 3.4 % of variance in suicidality (pseudo- elevation in posttraumatic stress and panic symptoms at T2
r2 =0.034). and T3, respectively; SC fully mediated the effect of time on
these two trauma-related symptom outcomes. Furthermore, in
Well-Being We tested the degree to which SC mediated the a non-lagged analysis, greater levels of SC at T1, T2, and T3
effect of time on well-being. The model demonstrated no were related to reduced levels of suicidality at each time-point,
significant effects, neither for time nor for SC; a non-lagged respectively; SC fully mediated the effect of time on
model similarly demonstrated no effects of time nor SC. suicidality-symptoms. In addition, levels of SC at T1, T2
and T3 were related to lower levels of depressive symptoms
Incremental Validity of Self-Compassion beyond Disposition- at each time-point respectively. In this latter test, SC could not
a l M i n d f u l n e s s f o r Tr a u m a - R e l a t e d S y m p t o m s mediate the effect of time because no significant effect of time
Outcomes Finally, we tested the incremental effect of SC was observed with respect to depressive symptoms; though it
above and beyond the effect of dispositional mindfulness is notable that in the non-lagged model, the effect of time on
(MAAS) on all symptom outcomes. However, MAAS scores depressive symptoms neared significance (p=0.06) and SC
did not explain significant variance in any symptom outcome accounted for this effect, consistent with theorized mediation.
nor change the reported associations between SC scores and It is furthermore important to highlight that the observed
symptom outcomes. MAAS scores were thus omitted from effects of SC with respect to trauma-related psychopathology
the MLM models of mediation. symptoms over-time were all unchanged and remained signif-
icant above and beyond the (null) effects of dispositional
mindfulness.
Together, these novel longitudinal and multi-level media-
Discussion tional data indicate that SC may function as a malleable
protective factor with respect to transdiagnostic trauma-
Scholars have increasingly focused on the role(s) of SC for related symptoms (i.e., depression, panic, posttraumatic stress,
mental health broadly and for the development and mainte- suicidality) among an at-risk sample of youth recently ex-
nance of stress- and trauma-related psychopathology more posed to a potentially traumatic stressor. The findings are
specifically (MacBeth and Gumley 2012). Yet, despite prom- novel with respect to the SC and mental health literature
ising theory, empirical study of the putative protective prop- (Barnard & Curry 2011; Neff and McGehee 2010; MacBeth
erties of SC with respect to trauma-related resilience and and Gumley 2012; Vettese et al. 2011) as well as traumatic-
recovery is limited. Accordingly, we tested whether SC pro- stress risk and protective factors literature (e.g., Bomyea et al.
spectively predicted reduced levels of posttraumatic stress, 2012; Brewin and Holmes 2003; Feldner et al. 2007).
depression, panic, and suicidality symptom, as well as well- Furthermore, these findings are consistent with the unique
being outcomes among an at-risk youth sample over the effects of SC on mixed anxiety-depression symptom severity
J Abnorm Child Psychol

observed above and beyond dispositional mindfulness (Van MAAS (dispositional mindfulness) and IDAS (symptoms)
Dam et al. 2011); though unlike Van Dam et al., no effect was were high and similar to past work; the SCS demonstrated
observed for dispositional mindfulness. However, unlike Van more modest levels of internal reliability. It is important to
Dam et al., the present study tested mediation, in a longitudi- highlight that though psychometrically acceptable, these mod-
nal design, and evaluated associations between these factors est levels of internal reliability of SCS result in a conservative
following traumatic stress. test of the role of SC in recovery post-trauma – notable in light
Furthermore, we observed significant elevation in posttrau- of the robust meditational effects of SC in the present study.
matic stress, panic, and suicidality symptoms, and a trend Moreover, it is important that future work evaluate the phe-
towards similar elevation in depressive symptoms, over the nomenological nature of SC in at-risk youth samples. We
6-months following traumatic stress exposure. These effects hypothesize that the present reliability data may mean that
are consistent with a large body of data documenting the some youth whom experience high levels of self-kindness or a
emergence of a range of forms of stress-related disorders sense of common humanity may also experience self-
among youth and at-risk youth in particular (Costello et al. judgment and a sense of isolation. This may well reflect the
2006; Grant et al. 2004). Contextualized in the larger body of complex nature of the phenomenology of self-compassion in
developmental psychopathology research related to stress and at-risk youth – an important question for future study that has
psychopathology among at-risk youth, these findings also to received limited attention to-date. Such future insights may
illustrate the importance of selective preventive or selective have important implications for clinical assessment and prac-
early intervention for youth who may be at elevated risk for tice. Fifth, symptom outcomes were not tested by means of
poor adjustment to trauma exposure. structured diagnostic data. However, use of the IDAS as the
The present study is limited in a number of respects rele- primary outcome measure in the present investigation may
vant to future work on SC and traumatic stress recovery. First, also be construed as a psychometric strength of the investiga-
the study was conducted in Israel, among a convenience tion. Indeed, the IDAS was developed to measure multiple
sample of youth at-risk. Future research sampling a more concurrent symptom outcomes while maximally
diverse population is important for further testing the gener- distinguishing between individual differences in these vari-
alizability of the present findings with respect to at-risk youth ables (Watson et al. 2007). Furthermore, outcomes were lim-
from a probability- or community sample. Second, measure- ited to internalizing symptoms – the role of self-compassion as
ment methods were exclusively retrospective and self-report. a protective factor for the development of externalizing symp-
Evaluation of SC using multi-method experimental measures, toms may be an important extension of the present findings
such as behavioral or experience sampling methods, are need- (Carrion et al. 2002; Saigh et al. 2002).
ed. Third, self-compassion was measured following exposure Sixth, though novel with respect to the temporal proximity
to the PTE. This of course does not permit us to test the to trauma exposure and at-risk status of the sampled youth, the
possible effect of the traumatic stress exposure on self- sample size was modest. Larger-scale future investigation is
compassion levels in the initial days following exposure. To important. Nevertheless, the longitudinal design and use of
do so, investigators would need to identify youth at-risk prior MLM significantly improved the statistical power of the anal-
to trauma exposure and to measure self-compassion or other yses (Preacher and Kelley 2011; Singer and Willett 2003).
protective factors in the days/weeks immediately prior to the Seventh, we were not able to directly test whether SC moder-
PTE. Doing so however is not readily feasible – unless inves- ates the effect of type or degree of trauma exposure severity on
tigators would focus on identification of youth at-risk of symptom recovery – due to the sample’s universal exposure to
trauma exposure – de facto changing the nature of the popu- the traumatic event. Future work in a sample variably exposed
lation and limiting the context for the study of SC and psy- to traumatic stress may permit such a test of moderation.
chopathology recovery post-trauma. Furthermore, clinically, Finally, though longitudinal, the study was observational.
were self-compassion measured and clinically targeted in a Participants were not randomized to a preventive intervention
future selective preventive or early intervention among at-risk targeting self-compassion. This design thus does not permit
youth, it would be similarly measured following PTE – as in making causal inferences regarding the observed longitudinal
the present study. Accordingly, knowledge of the role of self- associations between SC and trauma-related symptoms.
compassion in the days and weeks following trauma exposure In summary, we report novel data with respect to the
for symptom levels over time is important to guide future protective function(s) of SC for trauma-related adjustment
clinical research and practice. and psychopathology vulnerability among at-risk youth. The
Fourth, internal reliability of the studied measures ranged study has a number of notable strengths, including the unique
from acceptable to excellent; despite the socio-cultural differ- at-risk youth sample, recent exposure to a potentially traumat-
ences in this relative to majority-group samples in past work, ic stressor, longitudinal design, high prospective retention
there was no overall problem with the reliability of partici- rates, lagged multi-level mediation analyses, and
pants’ self-report. Specifically, levels of reliability on the psychometrically-sound measurement of symptom outcomes.
 J Abnorm Child Psychol

The findings may inform the development and investigation and theoretically derived future directions. Behavior Modification,
31, 80–116. doi:10.1177/0145445506295057.
of a novel trauma-related selective-prevention or early inter-
Fredrickson, B. L., Cohn, M. A., Coffey, K. A., Pek, J., & Finkel, S. M.
vention targeting SC among at-risk youth. (2008). Open hearts build lives: positive emotions, induced through
loving-kindness meditation, build consequential personal resources.
Acknowledgement Dr. Bernstein recognizes the funding support from Journal of Personality and Social Psychology, 95, 1045–1062.
the Israeli Council for Higher Education Yigal Alon Fellowship, the Geisinger, K. F. (1994). Cross-cultural normative assessment: translation
European Union FP-7 Marie Curie Fellowship International Reintegra- and adaptation issues influencing the normative interpretation of
tion Grant, Psychology Beyond Borders Mission Award, Israel Science assessment instruments. Psychological Assessment, 6, 304–312.
Foundation, the University of Haifa Research Authority Exploratory Gilbert, P., & Irons, C. (2004). A pilot exploration of the use of compas-
Grant, and the Rothschild-Caesarea Foundation’s Returning Scientists sionate images in a group of self-critical people. Memory, 12, 507–
Project at the University of Haifa. Mr. Zvielli recognizes the support from 516.
the University of Haifa President’s Doctoral Fellowship Program. Gilbert, P., & Procter, S. (2006). Compassionate mind training for people
with high shame and self-criticism: overview and pilot study of a
Conflict of Interest The authors declare that they have no conflict of group therapy approach. Clinical Psychology & Psychotherapy, 13,
interest. 353–379.
Grant, K. E., Compas, B. E., Thurm, A. E., McMahon, S. D., & Gipson,
P. Y. (2004). Stressors and child and adolescent psychopathology:
measurement issues and prospective effects. Journal of Clinical
References Child and Adolescent Psychology, 33, 412–425.
Hayes, S. C., Villatte, M., Levin, M., & Hildebrandt, M. (2011). Open,
aware, and active: contextual approaches as an emerging trend in the
Allen, A. B., & Leary, M. R. (2010). Self‐Compassion, stress, and coping. behavioral and cognitive therapies. Annual Review of Clinical
Social and Personality Psychology Compass, 4(2), 107–118. Psychology, 7, 141–168.
Barnard, L. K., & Curry, J. F. (2011). Self-compassion: conceptualizations, Hofmann, S. G., Grossman, P., & Hinton, D. E. (2011). Loving-kindness
correlates, & interventions. Review of General Psychology, 15(4), and compassion meditation: potential for psychological interven-
289. tions. Clinical Psychology Review, 31, 1126–1132.
Beaumont, E. A., Galpin, A. J., & Jenkins, P. E. (2012). ‘Being kinder to Hutcherson, C. A., Seppala, E. M., & Gross, J. J. (2008). Loving-
myself’: A prospective comparative study, exploring post-trauma kindness meditation increases social connectedness. Emotion, 8,
therapy outcome measures, for two groups of clients, receiving 720–724.
either cognitive behaviour therapy or cognitive behaviour therapy Kenny, D. A., Korchmaros, J. D., & Bolger, N. (2003). Lower level
and compassionate mind training. Counselling Psychology Review, mediation in multilevel models. Psychological Methods, 8, 115–
27(1), 31–43. 128.
Bomyea, J., Risbrough, V., & Lang, A. J. (2012). A consideration of Kraus, S., & Sears, S. (2009). Measuring the immeasurables: develop-
select pre-trauma factors as key vulnerabilities in PTSD. Clinical ment and initial validation of the Self-Other Four Immeasurables
Psychology Review, 32, 630–641. (SOFI) scale based on Buddhist teachings on loving kindness,
Brewin, C. R., Andrews, B., & Valentine, J. D. (2000). Meta-analysis of compassion, joy, and equanimity. Social Indicators Research, 92,
risk factors for posttraumatic stress disorder in trauma-exposed 169–181.
adults. Journal of Consulting and Clinical Psychology, 68, 748–766. Kross, E., & Ayduk, O. (2011). Making meaning out of negative expe-
Brewin, C. R., & Holmes, E. A. (2003). Psychological theories of riences by self-distancing. Current Directions in Psychological
posttraumatic stress disorder. Clinical Psychology Review, 23, Science, 20, 187–191.
339–376. Leary, M. R., Tate, E. B., Adams, C. E., Batts Allen, A., & Hancock, J.
Bronner, E. (2010). Facing fires, Israel appeals for help. The New York (2007). Self-compassion and reactions to unpleasant self-relevant
Times, pp. A5. events: the implications of treating oneself kindly. Journal of
Brown, K.W., Ryan, R.M. (2003). The benefits of being present: Personality and Social Psychology, 92, 887–904.
Mindfulness and its role in psychological well-being. Journal of Lutz, A., Brefczynski-Lewis, J., Johnstone, T., & Davidson, R. J. (2008).
Personality and Social Psychology, 822–848. doi:10.1037/0022- Regulation of the neural circuitry of emotion by compassion med-
3514.84.4.822. itation: effects of meditative expertise. PLoS ONE, 3, e1897.
Carrion, V. G., Weems, C. F., Ray, R., & Reiss, A. L. (2002). Toward an Lutz, A., Greischar, L. L., Rawlings, N. B., Ricard, M., & Davidson, R. J.
empirical definition of pediatric PTSD: the phenomenology of (2004). Long-term meditators self-induce high-amplitude gamma
PTSD symptoms in youth. Journal of the American Academy of synchrony during mental practice. Proceedings of the National
Child and Adolescent Psychiatry, 41, 166–173. Academy of Sciences of the United States of America, 101, 16369–
Costello, E. J., Foley, D. L., & Angold, A. (2006). 10-year research 16373.
update review: the epidemiology of child and adolescent psychiatric MacBeth, A., & Gumley, A. (2012). Exploring compassion: a meta-
disorders: II. Developmental epidemiology. Journal of the American analysis of the association between self-compassion and psychopa-
Academy of Child and Adolescent Psychiatry, 45, 8–25. thology. Clinical Psychology Review, 32, 545–552.
Davidson, R. J., & Harrington, A. (2001). Visions of compassion: Western Neff, K. D. (2003a). The development and validation of a scale to
scientists and Tibetan Buddhists examine human nature. USA: measure self-compassion. Self and Identity, 2, 223–250.
Oxford University Press. Neff, K. D. (2003b). Self-compassion: an alternative conceptualization of
Elwood, L. S., Hahn, K. S., Olatunji, B. O., & Williams, N. L. (2009). a healthy attitude toward oneself. Self and Identity, 2, 85–101.
Cognitive vulnerabilities to the development of PTSD: a review of Neff, K. D., Kirkpatrick, K. L., & Rude, S. S. (2007). Self-compassion
four vulnerabilities and the proposal of an integrative vulnerability and adaptive psychological functioning. Journal of Research in
model. Clinical Psychology Review, 29, 87–100. doi:10.1016/j.cpr. Personality, 41, 139–154.
2008.10.002. Neff, K. D., & McGehee, P. (2010). Self-compassion and psychological
Feldner, M. T., Monson, C. M., & Friedman, M. J. (2007). A critical resilience among adolescents and young adults. Self and Identity, 9,
analysis of approaches to targeted PTSD prevention: current status 225–240.
J Abnorm Child Psychol

Neff, K. D., Pisitsungkagarn, K., & Hsieh, Y. P. (2008). Self- to depression. The Journal of Positive Psychology, 5(5), 377–
compassion and self-construal in the United States, Thailand, 389.
and Taiwan. Journal of Cross-Cultural Psychology, 39, 267–285. Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis:
Ozawa-de Silva, B. R., Dodson-Lavelle, B., Raison, C. L., & Negi, L. T. Modeling change and event occurrence: Modeling change and
(2012). Compassion and ethics: scientific and practical approaches event occurrence. USA: Oxford University Press.
to the cultivation of compassion as a foundation for ethical subjec- Tasca, G. A., Illing, V., Joyce, A. S., & Ogrodniczuk, J. S. (2009). Three-
tivity and well-being. Journal of Healthcare, Science and the level multilevel growth models for nested change data: a guide
Humanities, 2, 145–161. for group treatment researchers. Psychotherapy Research, 19,
Pace, T. W. W., Negi, L. T., Adame, D. D., Cole, S. P., Sivilli, T. I., 453–461.
Brown, T. D., & Raison, C. L. (2009). Effect of compassion Tanaka, M., Wekerle, C., Schmuck, M. L., & Paglia-Boak, A. (2011). The
meditation on neuroendocrine, innate immune and behavioral linkages among childhood maltreatment, adolescent mental health,
responses to psychosocial stress. Psychoneuroendocrinology, and self-compassion in child welfare adolescents. Child Abuse &
34, 87–98. Neglect, 35, 887–898.
Preacher, K. J., & Kelley, K. (2011). Effect size measures for mediation Thompson, B. L., & Waltz, J. (2008). Self-compassion and PTSD symp-
models: quantitative strategies for communicating indirect effects. tom severity. Journal of Traumatic Stress, 21, 556–558.
Psychological Methods, 16(2), 93. Tirch, D. D. (2010). Mindfulness as a context for the cultivation of
Raes, F. (2010). Rumination and worry as mediators of the relationship compassion. International Journal of Cognitive Therapy, 3, 113–
between self-compassion and depression and anxiety. Personality 123.
and Individual Differences, 48, 757–761. Van Dam, N. T., Sheppard, S. C., Forsyth, J. P., & Earleywine, M. (2011).
Raes, F. (2011). The effect of self-compassion on the development Self-compassion is a better predictor than mindfulness of symptom
of depression symptoms in a non-clinical sample. Mindfulness, severity and quality of life in mixed anxiety and depression. Journal
2, 33–36. of Anxiety Disorders, 25, 123–130.
Reddy, S. D., Negi, L. T., Dodson-Lavelle, B., Ozawa-de Silva, B., Pace, Vettese, L. C., Dyer, C. E., Li, W. L., & Wekerle, C. (2011). Does self-
T. W. W., Cole, S. P., & Craighead, L. W. (2012). Cognitive-based compassion mitigate the association between childhood maltreat-
compassion training: a promising prevention strategy for at-risk ment and later emotion regulation difficulties? A preliminary inves-
adolescents. Journal of Child and Family Studies, 22, 219–230. tigation. International Journal of Mental Health and Addiction, 9,
Saigh, P. A., Yasik, A. E., Oberfield, R. A., Halamandaris, P. V., & 480–491.
McHugh, M. (2002). An analysis of the internalizing and external- Watson, D., O’Hara, M. W., Simms, L. J., Kotov, R., Chmielewski, M.,
izing behaviors of traumatized urban youth with and without PTSD. McDade-Montez, E. A., & Stuart, S. (2007). Development and
Journal of Abnormal Psychology, 111, 462. validation of the inventory of depression and anxiety symptoms
Shapira, L. B., & Mongrain, M. (2010). The benefits of self- (IDAS). Psychological Assessment, 19, 253–268. doi:10.1037/
compassion and optimism exercises for individuals vulnerable 1040-3590.19.3.253.