Maternal Emotion Socialization of Adolescent Girls Engaging in Non-Suicidal Self-Injury

Abstract

Non-suicidal self injury (NSSI) is a transdiagnostic maladaptive behavior that is highly prevalent in adolescence. A greater understanding of the mechanisms leading to NSSI is needed to guide the development of prevention efforts. The current study examined the relationship between maternal socialization of emotions and NSSI behaviors in their children. Female adolescents (N = 90, 12–17 years old) who demonstrated a range of NSSI lifetime episodes from none to very frequent were included in this sample. Maternal responses to their children’s displays of sadness, anger, and happiness were assessed. Principal components analysis was used to categorize items into supportive and unsupportive maternal emotion socialization responses for the three emotions. Adolescents whose mothers reported less supportive maternal responses to child’s expressions of sadness and anger had more lifetime NSSI episodes. Many of these patterns remained when follow-up analyses considered an extreme group approach (e.g., high counts of NSSI versus no NSSI), when analyses focused on specific diagnostic subgroups (e.g., depression and anxiety), and to some extent (socialization of anger) when current NSSI was considered. While the cross-sectional study design prevents causal conclusions, transactional theories raise the possibility that mothers’ emotion socialization may impact offspring NSSI and offspring engagement in NSSI may result in mothers altering their socialization practices to accommodate their child’s unique needs. Future research should employ longitudinal methodology to examine the time course, consider the role of emotion regulation as an explanatory mechanism, and consider intervention methods that may teach effective emotion socialization for parents.

Non-suicidal self-injury (NSSI) is defined as the act of harming oneself without the intent to die (Winchel & Stanley, 1991; Favazza, 1998). Nearly one in five adolescents have engaged in NSSI at least once in their lifetime and NSSI typically emerges during the adolescent years (Muehlenkamp, 2012). NSSI is associated with impaired functioning, reduced quality of life, frequent psychiatric hospitalizations, and elevates risk for suicide attempts and death by suicide (Apter et al., 2008; Nixon et al., 2008; Tang et al., 2011). NSSI is a transdiagnostic behavior that occurs across psychiatric and neurologic disorders (Apter et al., 2008; Tuisku et al., 2009) and also occurs in the absence of any disorders (Klonsky et al., 2003; Stanford & Jones, 2009).

Adolescents commonly grapple with the changes in cognition, emotion, behavior, and biology that transpire during this developmental period, they may experiment with a range of adaptive and maladaptive coping mechanisms to diminish discomfort and regulate emotions. In particular, adolescents who engage in NSSI commonly present with deficits in emotion regulation skills (Gratz & Roemer, 2008). Indeed one function of self-mutilating behavior is that it is automatically reinforcing (i.e., reinforcing to oneself), and thus serves to internally regulate emotion (Nock & Prinstein, 2004). While NSSI may serve to modulate negative emotions in the short term (Nock & Prinstein, 2004), it comes at a cost to long-term health and well-being. There is an urgent need to understand mechanisms leading to NSSI to guide intervention and prevention efforts.

Parents play an instrumental role in the development of emotion regulation in youth (Halberstadt, 1991). During normative development, children learn to express, understand, and regulate both positive and negative emotions through the process of emotion socialization (ES; Denham et al., 2009; Eisenberg et al., 1998), which continues to play a critical role into adolescence (Thompson et al., 2008). Parental ES occurs through three main pathways entailing both direct and indirect processes: modeling (parents’ expression of their own emotions), emotion coaching (emotion-centered discussions between parents and children), and contingent responding (parents’ real-time responses to children’s emotional expression; Denham, 1997; Eisenberg et al., 1998; Gottman et al., 1997).

Research links chronic parental invalidation with vulnerability in offspring to psychopathology. Emotionally invalidating family environments have been associated with NSSI among psychiatrically hospitalized adolescents, particularly for girls (Adrian et al., 2011). According to biosocial theories related to NSSI, early parental invalidation of a child’s emotional experience is a core risk factor for emotion difficulties typically seen in borderline personality disorder, which often presents with NSSI (Linehan, 1993; Crowell, 2009). This pattern of invalidation may also be applicable to ES. Parental neglect of negative emotions is linked with general psychopathology (Garside & Klimes-Dougan, 2002; Klimes-Dougan et al., 2007; Wright et al., 2009). Additionally, other unsupportive methods of parental ES to negative emotion, such as discouraging the expression of sadness by shaming the child, may contribute to distress and adversely influence youth’s ability to effectively regulate negative emotion (Klimes-Dougan et al., 2007). However, in the only study to date examining parental ES as it relates to NSSI, this pattern was not found. Specifically, Buckholdt and colleagues (2009) failed to find evidence that university undergraduate students’ recollections of unsupportive parental response to their sadness (such ignoring, neglecting, or punishing) was associated with NSSI severity, although power limitations were noted (e.g., only 27 participants had a lifetime history of NSSI).

An environment fostering acceptance and open discussion of emotions may serve as a protective factor against NSSI. Katz and colleagues (1999) described supportive responses to a child’s emotional expression as being “nestled in the web of positive parenting” (p. 142). Additionally, specific methods through which a parent scaffolds their child’s emotional repertoire may be critical to understanding mechanisms of NSSI risk. Much of the existing ES literature has sought to clarify ES of negative affect, specifically whether a parent encourages the child to process and regulate negative emotions. Supportive ES responses are tied to improved general emotion regulation ability (Shaffer et al., 2012) and lower psychopathology in adolescents (e.g., Brand & Klimes-Dougan, 2010). Importantly, in the previously mentioned paper by Buckholdt and colleagues (2009), parental supportive responses to participants’ sadness were inversely associated with NSSI. Specifically, young adults who engaged in NSSI indicated that their parent were less likely to reinforce or override their feelings of sadness, implicating supportive ES as a potential protective factor against the onset and development of NSSI.

The recent focus on the socialization of positive emotions is an important advance in the field. Regulation of positive emotions, such as happiness and joy, is an additional key factor in the development and maintenance of psychopathology (e.g., Werner-Seidler et al., 2013). Fredrickson’s “broaden and build” model posits that experiences of positive emotion form an “upward spiral” through which enduring emotion regulation resources are built (Fredrickson, 2004). Recent investigations suggest that parental positive affect is associated with child positive affect, but dampening of positive emotion and failure to reinforce positive behavior has been linked to clinical depression or other psychological problems (Katz et al., 2014; Gerhardt et al., 2019; Ramakrishnan et al., 2019). To date, no research has addressed parental socialization of positive emotions in the context of NSSI.

The pivotal role of parental ES in the development of emotion regulation underscores the importance of examining this process in adolescents with NSSI (Sanders et al., 2015). The current study examined supportive and unsupportive maternal ES in adolescent females on a spectrum of NSSI severity, from no reported history of NSSI engagement to very frequent NSSI engagement. We examined maternal socialization of both negative (sadness and anger) and positive (happiness) emotion in order to more thoroughly address gaps in the literature related to risks and processes of NSSI. We hypothesized that unsupportive responses to sadness, anger, and happiness would be associated with higher levels of lifetime NSSI, and that supportive responses to these emotions would be associated with lower levels of lifetime NSSI, even when controlling for a range of possibly intervening factors (e.g., participant age, maternal depression, youth depressive symptoms).

Method

Participants

Participants included 90 adolescents, all of whom were assigned female sex at birth. Participants were between the ages of 12–17 (M = 14.27, SD = 1.26) and were enrolled in a larger longitudinal investigation designed to use a multilevel approach to understand domains of functioning in adolescents with NSSI. Data collection for this longitudinal study is ongoing; the current project utilizes preliminary data from the baseline series of assessments.

Primary inclusion factors for the study were age, sex, and menarcheal status. The study oversampled for NSSI behaviors and participants represent a range of severity in NSSI engagement. Participants were excluded for Magnetic Resonance Imaging contraindications, history of developmental delay, head trauma, major medical illness, current substance abuse disorder, and/or lifetime primary bipolar or psychotic disorder. Because mothers and fathers significantly differ in how they use ES strategies in that mothers are generally more extensively involved in ES for sadness (e.g., fathers tend towards more punitive and neglectful strategies (Klimes-Dougan et al., 2007), the present study opted to specifically examine maternal ES, and thus only participants that had mother self-report data for the Emotions as a Child Negative and Positive Scales were included in this analysis. Participants whose fathers (n = 12) and grandmothers (n = 2) provided ES reports were excluded for this report.

Procedure

Participants were recruited from the greater metropolitan region of a city in the midwestern United States through flyers posted around the university and local community, digital marketing, and in- and outpatient clinics from local hospitals. This study was approved by the University of Minnesota. After completing a phone screen which reviewed initial inclusion/exclusion criteria, participants and their legal guardians attended an initial session which began with an informed written consent and assent process, followed by diagnostic assessment, self-report questionnaires, and intelligence testing. If diagnostic assessment confirmed eligibility, participants then completed two additional testing sessions, typically on subsequent days. These sessions involved the administration of the Trier Social Stress Test (Kirschbaum et al., 1993), behavioral computer tasks, and a functional Magnetic Resonance Imaging (fMRI) scanning session, visits which were not part of the current investigation. Participants were monetarily compensated for their participation. The current project only includes data from the initial visit of the baseline study session.

Measures

Demographics and Psychological Functioning.

Data on participant age, race/ethnicity, and gross family income were gathered via mother-report questionnaires. Participants and guardians completed the Kiddie Schedule for Affective Disorders and Schizophrenia-Present and Lifetime Version (KSADS-PL; Kaufman et al., 1997), a semi-structured diagnostic interview that assesses for Axis I diagnoses from the DSM-5. Approximately the first third completed the paper version (KSADS-PL) and two thirds completed the online version (the KSADS-COMP). Participants and mothers were interviewed separately by trained interviewers, and diagnoses were formed based on consensus of the two reports, using diagnoses generated from the online administration and through supervision with a team of licensed clinical psychologists. All interviews were conducted under the team’s supervision. The number of youth mental health diagnoses were based on these diagnoses. The mother report of both versions of the KSADS also inquired about a family history for major psychiatric disorders for first and second degree relatives. The estimate of maternal depression diagnosis was based on this parent’s response on the KSADS.

The Beck Depression Inventory - Revised (BDI-II) is a 21-item self-report inventory, designed to examine the severity of depressive symptoms in the past two weeks for those aged 13 to 80 (Beck et al., 1996). Each item is rated on a 4-point Likert scale, ranging from 0 (least severe) to 3 (most severe). The total scores are additive, and range from 0 to 63. The inventory has been shown to have good internal consistency in nonclinical adolescent populations (α = 0.92; Osman et al., 2008) and clinical adolescent populations (α = 0.93; Osman et al., 2004). The youths’ current symptom severity was based on the BDI-II.

Non-Suicidal Self-Injury.

The Self-Injurious Thoughts and Behaviors Interview (SITBI) is a structured interview that assesses multiple domains of self-injurious thoughts and behaviors, including suicide ideation, suicide plans, suicide gestures, suicide attempts, and non-suicidal self-injury (Nock et al., 2007). The NSSI subsection of the SITBI includes questions about NSSI frequency in the past week, month, year, and lifetime, age of onset, most recent NSSI episode, and type of self-harm. The focus on this study was on the frequency of lifetime NSSI episodes; “How many times in your life have you engaged in NSSI?” The assumption was that estimates would be similar to the preceding question on the SITBI pertaining to suicide thoughts and attempts. Here, inquiries were made about “separate” times and participants were encouraged to give their best estimate. The SITBI has strong interrater reliability (Cohen’s κ = .99). The test-retest reliability with a 6 month interval for the lifetime frequency of NSSI was on average Cohen’s κ = .71 (Nock et al., 2007). There was considerable variability for this index, but to minimize the chance that results would be due to outliers, values greater than 3 standard deviations above the mean were winsorized to correct for random sampling error and skew (n = 1). The assumption was that both the lifetime and past year measures of NSSI would be an estimate. Although more recent intervals were also assessed with the SITBI (e.g., past month), these were not considered here. However, follow-up analyses did consider the past year frequency of NSSI episodes, an index with a more constricted range, but nevertheless an index that was potentially less susceptible to memory decay. Follow-up analyses also used an extreme group approach where maternal ES was assessed for those within the highest and lowest quartiles of NSSI.

Emotion Socialization.

The Emotions as a Child Negative and Positive Scales, Version 2 (EAC2; Garside & Klimes-Dougan, 2002; Ramakrishnan et al., 2019) was based on Malatesta-Magai’s original Emotions as a Child Scale (Magai, 1996). This version focuses on parent reports (similar to Klimes-Dougan et al., 2007) and includes scales for sadness, anger, and happiness. Mothers rated items that corresponded with one of five ES strategies; three items represented each ES strategy, resulting in 45 items rated for each respondent. Mothers were asked to rate how typical each response was on a five-point scale (1 = not at all typical, 3 = somewhat typical, 5 = very typical). The five ES strategies assessed include: reward (providing comfort, empathizing, and problem solving); punish (conveying disapproval or making fun of a child); override (dismissive or distracting parental behaviors); neglect (parental behaviors that disregard a child’s emotional expression); and magnify (parental behaviors that match the child’s expression, e.g. a parent expressing happiness when their child is happy).

Consistent with standard scoring procedures, items were averaged to form emotion-specific scores for each response strategy (Klimes-Dougan et. al, 2007). While youth reports have typically been the focus of research, evidence of validity and reliability for the anger and sadness portions of this parental response scale have been previously reported (Klimes-Dougan et al., 2001; Klimes-Dougan et al., 2007). Parental responses to happiness from the EAC have not been previously investigated.

To reduce the number of ES variables and minimize multiple comparisons, three principal components analyses (PCAs) with oblique (direct oblimin) rotation were conducted. PCAs of the five response subscales were conducted separately for sadness, anger, and happiness to account for differences in the functional significance of specific responses to discrete emotions (e.g., magnification of anger versus happiness). Solutions were selected based on eigenvalues greater than one and visual inspection of the scree plot.

The PCA of responses to sadness yielded two components. The first component, representing unsupportive responses, as defined by punish (loading = 0.75), neglect (loading = 0.82), and magnify (loading = 0.73) responses, and accounted for 42.72% of the variance. The second component, representing supportive responses, included reward (loading = 0.77) and override (loading = 0.72) responses, and accounted for an additional 25.09% of the variance.

The PCA of responses to anger also yielded two components, but patterns of loading varied. The first component, again reflecting unsupportive responses, included punish (loading = 0.80), magnify (loading = 0.75), and override (loading = 0.83) responses, and accounted for 39.74% of the variance. The second component, representing supportive responses, was defined positively by reward (loading = 0.79) and negatively by neglect (loading = − 0.79). This component accounted for an additional 25.45% of the variance.

Finally, the PCA of responses to happiness yielded two components representing unsupportive and supportive responses, respectively. The unsupportive component was defined by punish (loading = 0.80), override (loading = 0.90), and neglect (loading = 0.65) responses, together accounting for 39.10% of the variance. The supportive component included reward (loading = 0.84) and magnify (loading = 0.83) responses and accounted for an additional 29.19% of the variance.

Emotion-specific component scores were calculated by averaging the relevant response indicators after reverse-coding neglect responses to anger. To test the internal reliability of the established component, we calculated the Cronbach’s alpha of each component using item-level data, and examined the total component alpha level if each item were to be removed.

Final Cronbach’s alpha for each factor are as follows: α = 0.75 for Sadness-Unsupportive, α =0.71 for Sadness-Supportive, α = .74 for Anger-Unsupportive, α = .66 for Anger-Supportive, α =.76 for Happiness-Supportive, and α = 0.74 for Happiness-Unsupportive. All items were included with the exception of one item from the Happy-Neglect subscale (“When your child is happy, how often do you not notice?”) that, when removed, increased the Cronbach’s alpha from 0.69 to 0.76.

Data analysis plan

Preliminary analyses were conducted to identify which demographic, clinical, and ES factors were significantly associated with lifetime NSSI. Pearson’s correlations were conducted to examine the associations between the six ES factors (supportive and unsupportive responses to sadness, anger, and happiness), participant age, race (white versus minority), family income, maternal depression, number of youth mental health diagnoses, current youth depression symptom severity, and lifetime and past year NSSI engagement.

ES factors that were significantly related to NSSI in bivariate analyses would be included as the focal predictor of NSSI in following Poisson regressions (Poisson Regression | R Data Analysis Examples, n.d.). Demographic and clinical factors that were significant in bivariate analyses were included in Poisson regression models as covariates. Following protocol from Cameron and Trivedi (2009), robust standard errors of the parameter estimates are reported below.

A series of follow-up analyses were conducted post-hoc to further examine the relationship between NSSI and ES. First, we ran all models with extreme groups (lifetime NSSI engagement was separated into the least and most severe quartiles), then within youth diagnostic subgroups (depressive disorders and anxiety disorders) to test if the associations between ES strategies and lifetime NSSI were replicated within more specific relatively well-represented internalizing diagnostic profiles. Finally, we examined all models looking at past year NSSI, as opposed to lifetime.

Results

Descriptive Information

Of our participants, 76% were non-Hispanic White, 8% Black/African American, 3% Asian American, 3% Native American, and 10% Hispanic/Latinx. While most of the analyses included NSSI count variables, study participants were included as controls if they had never engaged in non-suicidal self-injury in their lifetimes (n = 31, 34%) and included as non-controls if they had previously engaged in NSSI (n = 59, 66%). Descriptive data can be found in Table 1.

Table 1.

Total sample characteristics (N = 90)

Characteristic	n (%)
Age, mean (SD)	14.27 (1.26)
Race/Ethnicity
White and Non-Hispanic/Latinx	68 (75.56%)
Non-white or Hispanic/Latinx	22 (24.44%)
Hispanic/Latino	9 (10.00%)
African American/Black	7 (7.78%)
Asian	3 (3.33%)
American Indian/Alaska Native	3 (3.33%)
Gross Incomea
$10,000–$14,999	3 (3.30%)
$15,000–$24,999	1 (1.10%)
$25,000–$39,999	12 (13.33%)
$40,000–$59,999	4 (4.44%)
$60,000–$89,999	13 (14.44%)
$90,000–$179,999	37 (41.11%)
Over $180,000	19 (21.11%)
NSSI Engagement
No NSSI	31 (34.44%)
NSSI	59 (65.56%)
Mental Health Diagnosisa
No Diagnosis	23 (25.56%)
≥ 1 Mental health diagnosis	66 (73.33%)
No. of diagnoses/participant, mean (SD)	2.44 (1.98)
Depressive Disordersa	55 (61.80%)
Major Depressive Disorder	55 (61.80%)
Persistent Depressive Disorder/Dysthymia	7 (7.87%)
Anxiety Disordersa	45 (50.56%)
Generalized Anxiety Disorder	28 (31.46%)
Specific Phobia	19 (21.35%)
Social Anxiety Disorder	16 (17.98%)
Panic Disorder	8 (8.99%)
Separation Anxiety Disorder	7 (7.87%)
Agoraphobia	3 (3.37%)
Unspecified Anxiety Disorder	3 (3.37%)
Trauma and Stressor- Related Disorders,a	18 (20.22%)
Post-Traumatic Stress Disorder	16 (17.98%)
Other Specified Trauma and Stressor- Related Disorder	2 (2.25%)
Othersa
Attention Deficit Hyperactivity Disorder	21 (23.60%)
Eating and Feeding Disorders	9 (10.11%)
Encopresis/Enuresis	9 (10.11%)
Motor/Phonic Tic Disorder	5 (5.62%)
Oppositional Defiant Disorder	4 (4.49%)
Obsessive-Compulsive Disorder	3 (3.37%)
Trichotillomania	2 (2.25%)

^{a -}N = 89, 1 missing

Data are n (%) unless indicated otherwise

In this sample, 73% (n = 66) meet criteria for a mental health diagnosis. The most common diagnoses in the sample were depressive disorders (62%), with the average of total BDI-II scores being 14.29 (SD = 13.73). The number of clinical diagnoses participants met criteria for on the KSADS ranged from 0 to 8 (M = 2.44, SD = 1.98). Among those who engaged in NSSI, 86% had at least one mental health diagnosis, while 14% of those who had not engaged in NSSI had at least one mental health diagnosis. Number of lifetime NSSI episodes ranged in our sample from 0 to 404, with the average number of episodes being 37.19 (SD = 84.23) times. Of those who engaged in NSSI, onset occurred between the ages of 7 and 15 (M = 12.00, SD = 1.53) and the length of time engaging in NSSI was between 1 and 7 years (M = 3.12, SD = 1.53). Of the individuals in our sample who have engaged in NSSI, 86% endorsed cutting, 48% endorsed skin scraping, 37% endorsed purposeful self-hitting, and 37% endorsed picking at wounds. Some of the participants’ mothers (94% biological, 6% adoptive) endorsed having a history of depression (35%).

Bivariate associations

Results from bivariate analyses are reported in Table 2. Bivariate analyses revealed that lifetime NSSI engagement was positively associated with age (r(88) = .25, p = .02), maternal depression (r(87) = .27, p = .011), number of youth mental health diagnoses (r(87) =.24, p = .03), and total BDI-II score (r(77) = .34, p < .001). Further, lifetime NSSI engagement was associated negatively with the sad-supportive ES factor (r(88) = −.25, p = .02) and the angry-supportive ES factor (r(88) = −.27, p <.001). No significant relationship was found between lifetime NSSI engagement and sad-unsupportive, angry-unsupportive, happy-unsupportive, or happy-supportive ES.

Table 2.

Correlation matrix of demographic, clinical factors and emotion socialization factors (n = 90)

	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16
1. Lifetime NSSI	--
2. Past Year NSSI	.65***	--
3. Age	.25*	.17	--
4. Number of Diagnosed Disorders	.23*	.27**	.24*	--
5. Youth Depressive Symptom severitya	.34**	.42***	.22*	.55***	--
6. Maternal Depressionb	.27*	.08	.09	.11	.03	--
7. Depressive Disorderb	.14	.15	−.05	.32**	.38**	−.05	--
8. Anxiety Disorderb	.19	.19	.06	.52***	.18	.09	−.13	--
9. Gross Income	.12	.11	.06	−.05	−.04	−.26*	−.11	.07	--
10. Race/Ethnicity: Not white or hispanic/latinxb	−.06	−.13	−.02	−.11	.18	.02	.25*	−.22	.30**	--
11. Sad-Unsupportivec	.04	.19	.00	.17	.19	.07	.16	−.06	−.17	.14	--
12. Sad-Supportivec	.25*	−.17	−.06	−.04	−.19	.01	.03	−.01	.34**	.13	.30**	--
13. Angry-Unsupportivec	−.10	.01	−.05	.14	.06	−.04	.18	−.01	−.03	.21	.46***	.43***	--
14. Angry-Supportivec	−.27**	−.29**	−.26*	−.19	−.27*	−.11	−.07	−.03	.22*	.03	−.17	.47***	.02	--
15. Happy-Unsupportivec	−.05	.03	.04	−.01	.10	−.02	−.09	−.01	−.03	.11	.65***	.18	.37***	−.29**	--
16. Happy-Supportivec	.03	.00	−.01	−.12	−.17	−.15	.00	.11	−.09	.02	.09	.53***	.09	.33**	−.11	--
M	37.19	6.99	14.27	2.44	14.29	.35	.62	.51	7.36	.24	1.71	3.47	2.02	4.28	1.39	4.15
SD/Range	84.22	14.64	1.26	1.98	13.73	-	-	-	1.55	-	1.00–4.44	1.50–4.83	1.00–3.33	2.67–5.00	1.00–3.89	2.50–5.00

NSSI: non-suicidal self-injury.

^*p < .05;

^**p < .01;

^***p < .001

^{a –}Depressive symptom severity as determined by total Beck Depression Inventory-II score.

^{b -}Categorical variable (higher number “1” note a diagnosis of a depressive disorder/anxiety disorder/not-white or hispanic/latinx, “0” means no diagnosis of a depressive disorder/anxiety disorder/white).

^{c -}as determined by PCA analyses of EAC’s (Emotions as a Child Negative and Positive Scales, Version 2) item scores

Poisson Regression Models

Poisson regression was utilized to examine the association between lifetime NSSI frequency and ES. Four models were run in sequence for each ES factor controlling for: a) age, b) age and maternal depression, c) age, maternal depression, and number of youth mental health diagnoses, and d) age, maternal depression, and severity of current youth depressive symptoms. There were no significant models for a relationship between lifetime NSSI engagement and sad-unsupportive, angry-unsupportive, happy-unsupportive, or happy-supportive ES.

Sad-supportive.

Results revealed that when controlling for the effects of age, sad-supportive ES was significantly associated with lifetime NSSI engagement (β = − 0.81, p = .02) (Table 3). This association remained robust in follow-up tests controlling for maternal depression (β = − 0.87, p = .01), the number of youth mental health diagnoses (β = − 0.89, p = .003), and severity of youth depressive symptoms (β = − 0.72, p = .009).

Table 3.

Poisson regression modeling relation between sad-supportive emotion socialization (ES), angry-supportive ES and lifetime NSSI (n = 90)

	Lifetime NSSI
Sad-Supportive ES	β (SE)	95% CI
Model A
Age	0.47 (0.12)***, a	[0.22, 0.71]
Sad-Supportive ES	−0.81 (0.35)*	[−1.49, −0.12]
Model B
Age	0.40 (0.12)***, a	[0.17, 0.64]
Maternal Depression	1.10 (0.43)*	[0.25, 1.94]
Sad-Supportive ES	−0.87 (0.35)*	[−1.57, −0.18]
Model C
Age	0.32 (0.13)*	[0.07, 0.58]
Maternal Depression	1.07 (0.41)**	[0.27, 1.87]
# of Diagnoses	0.21 (0.10)* a	[−0.03, 0.37]
Sad-Supportive ES	−0.89 (0.30)**	[−1.47,−0.30]
Model D
Age	0.44 (0.14)**	[0.16, 0.72]
Maternal Depression	1.12 (0.37)**	[0.39, 1.85]
Depressive Sx Severityd	0.04 (0.02)**, c	[0.01, 0.07]
Sad-Supportive ES	−0.72 (0.28)**	[−1.27, −0.17]

Angry-Supportive ES	β (SE)	95% CI
Model A
Age	0.41 (0.13)***	[0.17, 0.66]
Angry-Supportive ES	−0.86 (0.41)*,b	[−1.65, −0.06]
Model B
Age	0.38 (0.11)***	[0.16, 0.61]
Maternal Depression	0.99 (0.46)*	[0.09, 1.89]
Angry-Supportive ES	−0.74 (0.38)*, a	[−1.49, −0.01]
Model C
Age	0.31 (0.13)*	[0.06, 0.55]
Maternal Depression	0.97 (0.44)*	[0.10, 1.84]
# of Diagnoses	0.22 (0.10)*, b	[−0.02, 0.41]
Angry-Supportive ES	−0.80 (0.29)**, b	[−1.37, −0.22]
Model D
Age	0.41 (0.14)**	[0.13, 0.68]
Maternal Depression	0.98 (0.41)*	[0.18, 1.78]
Depressive Sx Severityd	0.04 (0.02)*, b	[−0.00, 0.07]
Angry-Supportive ES	−0.61 (0.34)a	[−1.27, −0.06]

NSSI: non-suicidal self-injury, ES: emotion socialization, #: number, Sx: symptom.

^*p < .05;

^**p < .01;

^***p < .001

^{a -}Significant at p < .05 for past year NSSI;

^{b -}Significant at p < .01 for past year NSSI;

^{c -}Significant at p < .001 for past year NSSI

^{d -}Depressive symptom severity as determined by total Beck Depression Inventory-II score.

Angry-supportive.

Results revealed that when accounting for age, the relation between angry-supportive ES and lifetime NSSI engagement was significant (β = − 0.86, p = 0.04), and remained significant when controlling for maternal depression (β = − 0.75, p = 0.048), the number of youth mental health diagnoses (β = − 0.80, p = 0.006; Table 3). The relation between angry-supportive ES and lifetime NSSI trended towards significance when controlling for severity of youth depressive symptoms (β = − 0.61, p = .075).

Exploratory analyses.

Similar to the results from the poisson regressions, participants who were in the most severe lifetime NSSI quartile had lower sad-supportive ES scores (M = 3.22, SD = 0.44) than those in the least severe NSSI quartile (M = 3.56, SD = 0.50), t(24.55) = 2.47, p = 0.02). Also, the most severe lifetime NSSI quartile had significantly lower angry-supportive ES scores (M = 4.02, SD = .51) than those in the least severe NSSI quartile (M = 4.37, SD = 0.46), t(19.76) = 2.28, p = 0.03.

Additionally, post-hoc poisson regressions (controlling for age and number of youth mental health diagnoses) were conducted within depressive and anxiety subgroups to determine whether these associations between lifetime NSSI, sad-supportive and angry-supportive remained. Within the depression subgroup, lifetime NSSI remained significantly associated with sad-supportive (β = − 0.69, p = .04) and angry-supportive (β = − 0.85, p = .05) ES strategies (Table 4). In the anxiety subgroup, lifetime NSSI remained significantly associated with sad-supportive (β = − 0.81, p = 0.03), but not angry-supportive ES (β = − 0.66, p = .14) (Table 4).

Table 4.

Poisson regression modeling relation between sad-supportive emotion socialization (ES), angry-supportive ES, and lifetime non-suicidal self-jury (NSSI) in separate diagnostic subgroups.

	Lifetime NSSI
	Depressive Subgroup (n = 55)		Anxiety Subgroup (n = 45)
	β (SE)	95% CI	β (SE)	95% CI
Sad-Supportive ES
Age	0.35 (0.13)**	[0.09, 0.61]	0.43 (0.15)**	[0.14, 0.72]
# of Diagnoses	0.06 (0.13)	[−0.20, 0.32]	−0.02 (0.16)	[−0.33, 0.23]
Sad-Supportive ES	−0.69 (0.34)*	[−1.34, −0.03]	−0.81 (0.36)*	[−1.52, −0.10]
Angry-Supportive ES
Age	0.34 (0.14)*	[0.07, 0.61]	0.35 (0.14)*	[0.08, 0.62]
# of Diagnoses	0.02 (0.14)	[−0.24, 0.29]	−0.03 (0.17)	[−0.36, 0.30]
Angry-Supportive ES	−0.85 (0.44)	[−1.71, 0.01]	−0.66 (0.44)	[−1.52, −0.21]

NSSI: non-suicidal self-injury, #: number, ES: emotion socialization.

^*p < .05;

^**p < .01;

^***p < .001

To further understand the relationship between NSSI and ES, additional poisson regressions were run to examine the associations between these ES strategies with an alternative index of NSSI based on the number of episodes in the past year. Although poisson regressions revealed that sad-supportive ES was not significantly associated with past year NSSI engagement in any of the four models, angry-supportive ES was significantly associated with past year NSSI engagement in all models: controlling for age (β = − 0.93, p = .005), age and maternal depression (β = − 0.91, p = .01), age, maternal depression and number of youth disorders (β = − 0.90, p = .001), and age, maternal depression, and youth depressive symptom severity (β = − 0.63, p = .04).

Discussion

While the importance of preventing and treating adolescent NSSI is widely acknowledged (Miller & Smith, 2008), there remains limited work considering ways in which parents scaffold or undermine the adaptive use of emotions for adolescents who engage in NSSI. Notably, this study builds on current literature implicating NSSI as a transdiagnostic construct. The present investigation included a sample reflecting a range of DSM-V diagnoses and a full spectrum of NSSI severity to explore connections between ES and varying levels of NSSI. The central finding in the current study is that lower levels of supportive parental ES responses to negative emotions (sadness and anger) were linked with greater lifetime NSSI severity. These findings were robust in that similar patterns were generally noted when considering an extreme group approach, relevant diagnostic subgroupings (e.g., considering adolescents with depressive disorders), and past year NSSI (remained significant for anger but not consistently for sadness). Analyses confirm the robustness of these findings by showing that the results also largely remain significant when controlling for relevant covariates, including estimates of adolescent psychopathology and maternal depression. While causal theories were not tested, it is possible that failure to support discussions and processing of negative emotions may increase the likelihood and frequency of NSSI engagement in adolescence. The approach taken here leaves open the possibility that mothers may alter their strategies of addressing strong emotions when raising an adolescent who is engaging in NSSI (Linehan, 1993; Jobe-Shields et al., 2013). ES is reciprocal and dyadic in nature; theoretical models suggest that disruptive psychopathology may render mothers less likely to engage in supportive ES, implicating a bidirectional relationship in our results (Burke, Pardini, & Loeber, 2008).

The predictors of this study were exclusively relevant to socialization of negative emotion. Of note, the findings from the current investigation were robust regarding socialization of sadness. Our findings largely replicate and extend those reported by Buckholdt and colleagues (2009) who used a similar EAC scale but exclusively assess young adult’s perceptions of parental responses to sadness. Even with some critical differences between the studies’ methods, their findings virtually mirror the results in the current study. While our results were based on factor derived scales and implicate supportive responses, similar trends for our sample were noted for associations between NSSI and sadness-reward (r = − 0.2; p < .05) and sadness-override (r = − 0.19; p = .07) subscales. In addition, the factor loading for supportive responses to emotion are consistent with the model presented by Katz, Wilson, and Gottman (1999), wherein effective parenting is presented as a combination of allowance of emotional expression and scaffolding and redirection. Considering that our most robust results concerned socialization of sadness in lifetime NSSI, this is perhaps due to increased likelihood of NSSI to present in populations in which sadness is particularly salient or potentially a tendency for adolescents engaging in NSSI to turn upset and anger inward (Wolff et al., 2014).

The results of this study extend previous research (Buckholdt et al., 2009) suggesting that the link between NSSI and ES is not limited to the socialization of sadness, but instead may include responses to negative emotion more generally. Specifically, we found that maternal supportive responses to anger were inversely linked with both lifetime and past year NSSI frequency. Irrespective of whether the whole sample or a subgroup of those with internalizing pathology were considered (those diagnosed with depressive or and anxiety disorders), they showed the same pattern. Similar associations have been documented when considering links with internalizing and externalizing indexes of psychopathology (Klimes-Dougan et al., 2007). When parents fail to adequately support the expression of negative emotion, this may inadvertently encourage offspring to engage in experiential avoidance -- defined as the tendency to attempt regulation of negative thoughts, feelings, and situations through avoidance -- even when doing so causes problems or is maladaptive long-term (Hayes et al., 1999). Avoidance of negative emotional state has been theorized as a generalized psychiatric vulnerability and empirically validated as a mechanism underlying NSSI, among other psychiatric problems (Boeschen et al., 2001).

While supportive ES in particular may be crucial to child development, these findings may also reflect a broader constellation of other supportive parental practices including reduced warmth, nurturance, and attachment. For example, the current study is preceded by an expansive literature implicating a “buffering effect” of secure attachment in numerous variables relevant to psychopathology and general well-being (e.g., Karreman & Vingerhoets, 2012; Huebner & Thomas, 1995). Specific to NSSI, Baetens and colleagues (2015) found evidence for parental support as a buffer against bullying and other forms of victimization that may increase risk for the development of self-injurious behavior. In light of the current study’s results, mothers of adolescents who engage in NSSI may struggle to find appropriate strategies for interacting with their children who struggle with self-harm. Supportive parental responses to negative emotion may be just one component of parental behavior potentially showing a connection with the development and maintenance of NSSI.

Although this was the first study that examined a possible link between maternal emotion socialization of happiness and offspring NSSI, this investigation failed to find connections between ES responses to positive emotion and NSSI, even in the subgroup of participants diagnosed with depression. This is in contrast to literature implicating parental ES strategies aimed at savoring and dampening positive emotion in the development of clinical depression in offspring (Sanders et al., 2015; Katz et. al., 2014) and Frederickson (1998) who found that parental supportive ES of positive emotions is correlated with the child’s ability to employ positive affect. A possible reason for this discrepancy lies in the variability of depressive symptoms in our population, or a potential difference in the way adolescents with NSSI display positive affect.

Finally, the current investigation did not find a significant relationship between unsupportive ES responses and NSSI, like providing response contingencies that likely discourage the expression of emotion (neglect, punish). This may reflect limitations of maternal self-report for assessing ES. Mothers may be less likely to recall or endorse overtly unsupportive behavior than they would be to endorse lower levels of supportive behavior. Indeed, past research has shown evidence of a link between supportive parental responses for correspondence between parent and child reports, but this association has not been reported for punishing, neglecting, and magnifying parental ES practices (Klimes-Dougan et al., 2009). Not only may parental reports of unsupportive responses be less valid, these results would not likely extend to child-reported ES. Additionally, it is possible that more insightful mothers may be more likely to notice and report unsupportive responses. These response styles would in turn curtail the associations with NSSI. While generally the scale loadings for discrete emotions were similar to past work with other samples (e.g., override typically loads with reward for responses to sadness; Klimes-Dougan et al., 2009), there were some minor differences. For example, unsupportive and supportive emotion socialization were positively correlated (Table 1). Considering this atypical response style further, follow-up analyses were conducted using a person-centered approach to identify ES groups based on all 15 subscales. The results of the cluster analyses revealed support for a two-group model, one with higher endorsed ES responses across strategy and emotion; the other with lower ES responses across strategy and emotion. Interestingly, NSSI count was higher for youth when mothers were less involved in their children’s emotional lives (irrespective of endorsing supportive or unsupportive strategies), as compared to those more involved. Finally, another parent or caregiver in the home could be employing supportive ES and serving as a buffer against the onset or worsening of NSSI, making up for the shortcomings of the mother (Brand & Klimes-Dougan, 2010).

This study provides important foundational information about the association between maternal emotion socialization and NSSI, but some study limitations should be accounted for when considering next steps for this line of work. Measurement issues relevant to NSSI have plagued the field (Nixon & Heath, 2009). While the SITBI is well-validated and can provide detailed information about the frequency and level of severity of self-injury, some are likely to consider this information highly personal and may prefer to keep the information private (Boldero & Fallon, 1995). The results of this study extend previous research that has considered lifetime NSSI (Buckholdt et al., 2009), by assessing both lifetime and past year NSSI. The assumption is that these values constitute an estimation. For example, when participants indicated uncertainty regarding the total number of NSSI episodes they had engaged in within the context of the SITBI, they were encouraged to produce an estimate. Additionally, each index of NSSI counts has unique advantages and disadvantages. While variability of NSSI episodes in the sample is greater for lifetime than past year, lifetime recall of self-harm is subject to under-reporting bias (Mars et al., 2016). The findings that differed the most between lifetime and current NSSI were the associations with sadness emotion socialization, specifically when youths’ current depressive problems were controlled for in the model. Current depressive problems of youth may be more closely linked to the ways mothers respond to sadness. Another study strength is that NSSI was based on youth report, with ES ratings based on maternal report. This approach avoids possible inflation of association due to shared informant responses, parental responses tend to show a number of biases due to impression management, social desirability, lack of insight, etc. (Klimes-Dougan et al., 2007). Future research should also account for adolescents’ perspectives ES and for multiple caregivers, potentially including a teen report and measures specific to fathers. Indeed, mothers and fathers differ substantially in the socialization of their children. For example, fathers’ responses to emotions may be particularly relevant to transdiagnostic risk (Ramakrishnan et al., 2019).

In addition to internal validity, generalizability may be limited. The results may not apply to teen males with NSSI; fathers’ emotion socialization practices; individuals in other age groups; or individuals of other races, ethnicities, and socioeconomic status. Additionally, this study does not include information with regard to sexual orientation or gender identity, so the extent to which the results apply to adolescents who identify as sexual or gender minorities is unknown. Finally, it is possible that this study over-sampled for supportive, well-resourced families for which mothers were willing to participate in the larger three-visit study and were typically aware of their offspring’s NSSI. A line of research that will require more attention in the future is the role of family resources. Family resources may be viewed in light of parental health status. Past research has shown a link between maternal depression diagnosis and ES (Labella et al., 2020). Our estimate of maternal depression was also significantly correlated with ES and NSSI (Table 2) and when we included it as a covariate in the regression models, all findings remained significant (Table 3 and 4). Financial resources may also impact behavior. While our study sample was skewed to primarily well-resourced families, there was a subgroup of participants from low SES families. As noted in Table 2, family SES was inversely correlated with supportive maternal responses to anger and sadness. When models were rerun including family income grouping as a covariate, the links between ES and NSSI continued to show the same pattern but were no longer significant. Future research is needed to address these key study limitations. Finally, important next steps would ideally involve longitudinal data using cross-lagged designs which elucidate parental ES mechanisms in NSSI persistence and resolution.

This study serves to fill a gap in the literature concerning potential mechanisms of NSSI. Further research is needed to clarify the progression of experiences that takes place within the individual that sets the stage for NSSI. Exploration of a variety of possible mediators and moderators is pertinent to the understanding of parental ES and NSSI development and maintenance. The literature suggests that parental socialization of negative emotion plays a mediating role in emotion dysregulation in children (Sanders et al., 2015; Buckholdt et al., 2009). In her biopsychosocial model, Linehan (1993) advanced the idea that an individual’s predisposition to experience intense emotions may interfere with functioning when coupled with inadequate environmental opportunities for learning to manage emotions. The idea that inherited vulnerabilities, in conjunction with negative experiences, lead to engagement in NSSI has been further supported in primate models (Tiefenbacher et al., 2005). Crowell and colleagues (2009) applied these concepts more specifically to the etiology of NSSI, suggesting that emotion dysregulation fosters and maintains self-injury within hostile and invalidating social contexts. Another related avenue to examine in the future is the role of trauma in amplifying emotion dysregulation and risk for NSSI. In addition, while the literature delineating connections between childhood trauma and NSSI and related diagnoses is relatively expansive, correlations between childhood sexual trauma and NSSI are particularly consistent (e.g., Frazier et al., 2009; Herman et al., 1989).

The current findings underscore the importance of addressing both parenting and adolescent psychopathology in NSSI. These results have potential to inform intervention methods promoting mindful, sensitive parenting practices both for prevention of NSSI onset and for parents of children who already engage in NSSI. Supportive socialization is a dynamic, changing process which evolves as a child ages (Mirabile et al., 2016); future research should investigate and expand interventions which can be appropriately tailored to children’s individual needs and age, such as ES-based strategy Tuning in to Teens (TINT; Kehoe et al, 2013). In addition, investigations exploring therapeutic methods for families of adolescents with NSSI should expand on the current study in order to produce interventions specifically aimed at reducing NSSI severity, developing ES-based intervention strategies for parents, and enhancing existing treatment for families affected by NSSI.