Regulating Emotion and Identity by Narrating Harm

Abstract

This study examined how narration of harm experiences can regulate self and emotions in ways relevant to well-being. Participants (n = 88, 65% female) were asked to provide 6 narratives about instances when they were victims of harm and 6 narratives about instances when they were perpetrators of harm. Narratives were coded for extent of exploration, growth, damage conclusions and resolution. Participants drew damage conclusions more frequently in victim narratives and growth conclusions more frequently in perpetrator narratives. Both the type of experience (victim or perpetrator) and the way the experience was narrated (references to damage conclusions and resolution) predicted emotion and identity implications, which were, in turn, related to well-being. Implications for narrative approaches to self-regulation are discussed.

Regulating Emotion and Identity by Narrating Harm People narrate the majority of emotional experiences in their lives to others – often within the same day that those events occur (Pasupathi, McLean, & Weeks, 2009; Rimé, Finkenauer, Luminet, Zech, & Phillipot, 1998). Such narration likely serves to regulate emotions associated with the events (Nils & Rimé, 2012; Pasupathi, 2003; Pasupathi, Wainryb, Mansfield, & Bourne, under review; Skowronski, Gibbons, Vogl, & Walker, 2004), as well as to regulate the identity or self-implications those events may hold (Mansfield, Pasupathi, & McLean, under review; McLean & Pasupathi, 2011; McLean, Pasupathi, & Pals, 2007; McLean & Pratt, 2006; Pasupathi, Mansour, & Brubaker, 2007; Waters & Fivush, 2014; Weeks & Pasupathi, 2011). In the present study, we examined links between the way people narrate experiences of interpersonal harm and their emotions about the experience, and perceptions of the experience as identity central. We then examined the relationships between emotions and identity-centrality perceptions and measures of well-being.

The Story that Reduces Emotional Distress and Regulates Identity Implications

In constructing stories about challenging, important, and difficult experiences, people may explore various aspects of their experiences, draw conclusions about the meaning of those experiences, and create a sense of closure or resolution for the experiences (McAdams, 1996; McLean et al., 2007; Pasupathi, 2001). Positive and growth-promoting conclusions and resolution are linked to reductions in negative emotion (Pasupathi et al., under review), and are positively associated with more mature identities and higher levels of well-being (Lilgendahl & McAdams, 2011; McLean et al., 2007; Weeks & Pasupathi, 2011). Exploration is more closely connected to the processing of experiences rather than to a distinct outcome, and can have different implications for distress, versus broader indices of well-being. Some findings suggest that exploration of negative emotions may actually exacerbate negative emotion (Pasupathi et al., under review). A larger body of work shows relationships between higher levels of exploration, positive conclusions, and resolution in people's narratives, and indices of higher well-being and more advanced identity development (Bauer & McAdams, 2010; Fivush, Bohanek, & Marin, 2010; King & Patterson, 2000; Lilgendahl & McAdams, 2011; McAdams, Reynolds, Lewis, Patten, & Bowman, 2001; McLean et al., 2007).

Although a central proposition in narrative identity research is that the way people narrate experiences influences whether and how those events become identity-defining, this idea has not been tested. Much of the work reviewed above elicits events already deemed identity central (Blagov & Singer, 2004; McAdams, 1996; McAdams et al., 2001; McLean, 2005). Further, much of the above work takes a between-person approach, in which researchers connect between-person differences in narrative features to between-person differences in well-being, emotion, and identity. This emphasis on between-person differences and already important events precludes examining how variability in the way people narrate a variety of their own experiences relates to variability in their perception that events are identity central (McLean et al., 2007; Pasupathi et al., 2007). This is a core assumption for narrative identity researchers. So, a within-person approach to relationships between narrative features and indices of emotion and identity centrality would constitute a stronger test of this proposition.

Emotion and Identity Challenges Implicated in Narrating Interpersonal Harm

In the present studies, we focus on experiences of interpersonal harm. Such experiences entail negative emotions like anger, sadness, guilt, and shame, and challenge our views of ourselves and others as morally upright. Narrative construction of these experiences can help to regulate emotion and identity implications, but not always in adaptive ways (Pasupathi & Wainryb, 2010; Recchia, Wainryb, & Pasupathi, 2013). For example, experiences of harm can be narrated in ways that construct a sense of the potential for growth, increased understanding of ourselves and important others in our lives, and as imperfect but still fundamentally good human beings (Mansfield et al., 2010; Pasupathi & Wainryb, 2010; Wainryb & Pasupathi, 2010). But they can also be narrated in ways that ruminate on distress and create a sense of the self as bad or damaged (Baumeister, Stilman, & Wotman, 1990; Lilgendahl, McLean, Mansfield, 2013; Wainryb & Pasupathi, 2010; Wainryb, 2011).

The challenges and potential for experiences of harm, and the way that they are narrated, could also differ depending on the perspective of the narrator – whether the narrator was the victim or perpetrator of that harm (Baumeister et al., 1990; Schütz, 1999; Wainryb, Brehl, & Matwin, 2005). Compared to narratives about being the victim of harm, narratives about perpetrating harm are more complex, shift between the narrator's perspective and the victim's perspective more frequently, and consider multiple features of the experience, including the goals, beliefs, and intentions of all involved (Baumeister et al., 1990; Wainryb et al., 2005). Narratives about perpetrating harm may also seek to justify or otherwise mitigate the identity implications of the events (Baumeister et al., 1990). These findings, while not directly focused on exploration, growth, damage, and resolution, suggest that perpetrator narratives about harm will evince more exploration, and potentially a higher frequency of growth or damage meaning-making. Previous findings that perpetrator narratives tend to portray an event as over, while victim narratives from the same narrator tend to portray the event as having lasting negative consequences, suggest that perpetrator narratives will be more resolved, and that victim narratives may be more likely to be linked to damage conclusions.

Broadly, the narrative literature suggests that narrating experiences with more exploration, growth, and resolution is linked to higher well-being (see, e.g., McLean et al., 2007). There are at least two, non-mutually exclusive possibilities about how those links arise. One possibility is that in narrating experiences, people construct a sense of how central that event is to their identity in ways that affect well-being. For traumatic events, viewing the event as central to identity has been linked to increased distress and a higher rate of reported PTSD symptoms (Berntsen & Rubin, 2007; Greenhoot & Bunnell, 2009); while viewing an event as having engendered growth is linked to higher well-being, as noted above. A second possibility is that in narrating experiences, people reduce the distress they feel about the event both currently, and in lasting ways (Pasupathi et al., under review), such that reminders of the event are no longer as distressing.

Summary and Present Study

The present study involved assessing six victim and six perpetrator narratives from each participant for exploration, growth meanings, damage meanings, and resolution. In addition, we asked participants to rate each event for how they currently felt about it along basic (anger, sadness, fear, disgust) and moral (shame, guilt) negative emotions, and to rate the extent to which each event was central to their identity. We also asked participants to complete a battery of personality and well-being assessments. For the present study, we focus on the subset of those measures that addressed aspects of well-being (life satisfaction, eudaimonic well-being, self-compassion, and forgiveness).

This design allowed us to address several broad questions and specific hypotheses. First, we tested whether victim and perpetrator events differed in the way participants narrated them in terms of exploration, growth, damage, and resolution; we expected perpetrator events to be narrated with more exploration, growth, and resolution, and victim events to be narrated with more damage. Second, we asked how exploration, growth, damage, and resolution in the narratives related to participants emotional and identity-related judgments about the narrated events. We expected that exploration, growth, and damage might be related to judging events as more identity central, and that growth and damage would be related to decreased and increased negative emotions, respectively. We also expected resolution to be associated with decreased negative emotion. In a more exploratory vein, we asked whether these relations were moderated by the type of event – for example, people might engage in exploratory processing of perpetrator events in an effort to dismiss the identity-relevance of the event. We therefore also tested interactions between narrative qualities and the type of event (perpetrator or victim) in predicting emotion and identity. Last, we examined the relationship between well-being and aggregate measures of narrative features, identity and emotions. Based on previous work, we expected that more exploration, growth, and resolution in the narratives would be linked to higher well-being; analyses linking identity and emotion to well-being were exploratory.

Method

Participants

A collegiate sample was recruited from the undergraduate research participation pool at a major university in the western United States. Participants were 88 (65.50% female) students who received two hours of research participation credit in exchange for their responses. The sample was primarily European-American (79.30 %) and young (M age= 21.80, SD= 3.97). Four participants were excluded due to issues arising during data collection. The remaining participants were included in analyses provided they contributed at least 5 stories to the dataset (n = 2); a total of 932 stories and 84 participant were included in the present study. Based on rules of thumb for power in multi-level modelling (Kenny, Kashy, & Cook, 2006), this design provides us with adequate power detecting medium effects in models that include a maximum of 11 fixed effects and interactions.

Procedure

Participation entailed a single 2 hour long session. First, participants were asked to provide narratives about 12 experiences: 6 in which they were harmed by another individual (Please tell us about times when someone did or said something and you felt hurt by it) and 6 in which they harmed another individual (Please tell us about times when you did or said something and someone ended up feeling hurt by it). The order in which participants provided the two types of narratives was counterbalanced. Immediately after providing each narrative, participants were asked to rate that event for its identity centrality, and for the emotions they experienced while recalling the event. Following the elicitation of narratives, participants completed questionnaires (see below). The order of the questionnaires was also counterbalanced. Finally, participants were instructed to write about an “event that makes you feel really good”; this was done to ameliorate any distress participants experienced after recalling multiple negative experiences.

Measures

Satisfaction with Life Scale

This 5-item scale measures the degree to which one is satisfied with life (Diener, Emmons, Larsen, & Griffin, 1985). An example of an item from the scale is, “If I could live my life over, I would change almost nothing.” Responses were provided on a 1 (strongly disagree) to 7 (strongly agree) Likert scale. Scores were calculated by summing all of the items. The scale produced adequate reliability (α= .81).

Well-Being Scale

A shortened version of the original Ryff Well-Being Scale was used. The scale includes 18 items, 3 items representing each of the six dimensions: self-acceptance, environmental mastery, positive relations, purpose in life, personal growth and autonomy (Ryff & Keyes, 1995). An example of an item from this scale is, “I feel good when I think of what I've done in the past and what I hope to do in the future.” Responses were provided on a 1 (strongly disagree) to 6 (strongly agree) Likert scale. The six subscales produced adequate reliability (α= .81).

Self-Compassion Scale

We used a shortened version (12-items) of the original 26-item scale which measures the extent to which individuals have compassion for themselves (Neff, 2003a). In this scale, self-compassion is defined as openness to one's own suffering as well as caring for, kindness towards and non-judgment of the self (Neff, 2003b). An example of an item from this scale is, “When things are going badly for me, I see the difficulties as part of life that everyone goes through.” Responses were provided on a 1 (almost never) to 5 (almost always) Likert scale. This scale achieved adequate reliability (α= .89).

Forgiveness (Transgression-Related Interpersonal Motivations)

This 18 item scale measures the extent to which people respond to being the victim of harm with avoidance, vengefulness, or benevolence (McCullough & Hoty, 2002). We made a small adaptation to the instructions for this measure by instructing participants to think about how they respond in general, rather than having them think about a particular harm incident. An example of an item from this scale is, “Although he/she hurt me, I am putting the hurt aside so we can resume our relationship”. Responses were provided on a 1 (strongly disagree) to 5 (strongly agree) scale. Each of the 3 subscales achieved adequate reliability (avoidance α= .81, revenge α= .90, benevolence α= .88).

Composite Well-Being

As can be seen in Table 1, the well-being measures were relatively strongly inter-correlated (p<.001). Thus, we created a composite well-being score by standardizing and aggregating scores across the four well-being measures (see Waters & Fivush, 2014 for a similar approach).

Table 1. Intercorrelations among well-being measures.

Measure	1.	2.	3.	4.
1. Self-Compassion		.59***	.67***	.51***
2. Life Satisfaction			.65***	.59***
3. Ryff				.49***
4. TRIM

Note.

^*= p<.05,

^**= p<.01,

^***= p<.001

Emotion Ratings

Participants rated the anger, fear, sadness, disgust, guilt, and shame that the event currently elicits on 7-point Likert scales (1 = not at all, 7 = very). We used these ratings to create two composites, one for basic negative emotions (anger, sadness, disgust and fear, α= .77) and one for moral emotions (guilt and shame, α= .84). The creation of the composites was supported by exploratory principle components analysis with an oblimin rotation, which yields two factors accounting for 69.9 percent of the variance, and clear distinct loadings of basic negative emotions on one factor and moral negative emotions on the other. Although disgust is sometimes argued to have moral aspects (e.g., Rozin, 1996), in the present data, disgust clearly loaded on the basic negative emotion factor. Reported results do not differ when disgust is excluded. For perpetrator events, participants' reported average basic negative emotions at 2.5 (1.0) and moral emotions at 3.2 (1.1) on a 7-point scale. For victim events, these averages were 3.5 (1.3) nd 2.0 (0.9) respectively.

Identity Centrality Rating

Following each story participants were asked to rate the extent to which the event they described was central to their identity on a 7-point Likert scale from 1 (not central at all) to 7 (completely central). For perpetrator events, participants reported an average of 3.0 (1.3) on the centrality item, and for victim events, a slightly higher average of 3.5 (1.4).

Excluded Measures

The following measures were collected for the purpose of assessing individual differences in social and personality functioning in relation to narrating harm, but were excluded from the current study: Basic Needs Satisfaction in Life Questionnaire (Deci & Ryan, 2009), Big Five Inventory (John, Donahue, & Kentle, 1991), Experiences in Close Relationships Questionnaire (Wei, Russell, Mallinckrodt, & Vogel, 2007), Mechanisms of Moral Disengagement Questionnaire (Bandura, Barbaranelli, Caprara, & Pastorelli, 1996), Multidimensional Scale of Perceived Social Support (Zimet, Dahlem, Zimet, & Farley, 1988), Need for Closure Short Questionnaire (Roets & Van Hiel, 2011), and Self-Reported Executive Function (Buchanan et al., 2010). For the first event of each type, participants were also asked to report on the disclosure history for that experience and these data are also not included.

Coding

Narratives were coded for exploratory processing of positive and negative aspects of the event, growth conclusions, damage conclusions, and resolution (vs. rumination), using an adaptation from the Coding System for Components of Transformational Processing (Pals, 2006). We adjusted the original scheme from 5- to 3-point scales given our shorter narratives. Exploratory processing is reflected in narrative content that suggests the narrator has thought about the experience and its implications. Coders initially disagreed about exploration codes; in examining these disagreements, it appeared that coders were having difficulty with weighting exploration of positive and negative aspects of the events. Thus, we separated positive and negative exploration to improve reliability. Growth conclusions are indicated by statements suggesting positively perceived changes in the person's understanding of self and world. Damage conclusions were reflected by statements suggesting lasting negative consequences for the self. Rumination vs. resolution was scored from statements suggesting the event was in the past, versus statements suggesting the event still intruded on the person's current life. Positive and negative exploratory processing, growth conclusions, and damage conclusions were scored on a 0 to 2 scale with 0 indicating absence, 1 indicating some evidence of exploratory processing, and 2 indicating strong evidence of exploratory processing. Resolution/rumination was also coded on a three point scale where -1 was rumination, 0 was unclear, and 1 was resolved. Examples of the coding scheme are provided in Table 2.

Table 2. Descriptive statistics for perpetrator and victim narrative explorations.

	Perpetrator M(SD)	Victim M(SD)	Percentage of Narratives Rated as High in Characteristic	Percentage of Narratives Rated as Medium in Characteristic	Percentage of Narrative Rated as Low in Characteristic	Example
Positive Exploration	.07 (.23)	.06 (.12)	1.4 %	3.7 %	94.9 %	“She was looking out for the dignity of the family and was trying to help everyone when she hurt me.”
Negative Exploration	.35 (.39)	.36 (.41)	10.5 %	14.9 %	74.6 %	“She was worried about looking embarrassed in front of others and looked out for her own self-interest.”
Growth Conclusions	.08 (.25)	.02 (.07)	0.4 %	2.5 %	97.1 %	“I feel now that I can trust myself more…”
Damage Conclusions	.02 (.05)	.07 (.12)	0.5 %	3.3 %	96.2 %	“I feel now that I can't trust anyone and that anyone you know will let you down.”
Resolution	.00 (.20)	.00 (.13)	4.6 %	90.1 %	5.3 %	I was hurt by her actions, but I have since then moved on to a better partner and don't think about her at all.”

Note. Scales were 0-2 for all elements except for resolution which was on a -1 to 1 scale.

Three coders, including the first and fifth authors as well as an undergraduate coder blind to all hypotheses, first read an initial manual with examples and definitions. They then met weekly to discuss coding and to resolve disagreements until adequate reliability was obtained. After this threshold reliability, the fifth author and the undergraduate coder coded all of the data. To assess the maintenance of reliability over the course of coding, each dimension was coded independently by the fifth author, an undergraduate coder, and the first author for a randomly selected subset of the data (19 participants per dimension). Each dimension was reliability checked for different subsets of participants and narratives; a total of 35 participants' narratives, and a total of 391 stories were reliability scored for at least one of the dimensions (8 reliability-coded participants were missing some stories – 3 were missing 6 stories, 2 missing 9, 2 missing 10 and 1 missing 11). Percentage agreement was as follows: positive exploration, 80.4%; negative exploration, 89%; growth, 96%; damage, 89%; and resolution, 86%. Because some codes were extremely low in frequency, using Cohen's kappa is problematic (Andres & Marzo, 2004), but percentage agreements are not corrected for chance. To address this issue, we computed a coding-scheme-wise measure of agreement across all pairs of coder decisions for each dimension and story for 19 participants (total n = 399 stories; some participants were missing some stories). For the entire coding scheme, percentage agreement was 88%, and Cohen's kappa was .66, p < .001.

Results

Preliminary Analyses

Table 3 contains means and standard deviations for each narrative feature. The data reveal that narrative features varied widely in their frequency. Negative exploration occurred frequently and positive exploration was infrequent. Growth and damage conclusions also happened relatively infrequently, and most narratives were neither particularly resolved nor unresolved.

Table 3. Intercorrelations among narrative features for perpetrators and victims.

Narrative Element	1.	2.	3.	4.	5.
1. Positive Exploration	1	.31**	.38***	-.07	.31**
2. Negative Exploration	.28**	1	.09	.10	-.001
3. Growth Conclusions	.47***	.35**	1	-.07	.20
4. Damage Conclusions	.09	.24*	.25*	1	-.24*
5. Resolution	.12	-.05	-.04	-.03	1

Note.

^*= p<.05,

^**= p<.01,

^***= p<.001.

Perpetrator correlations above the diagonal and victim correlations below the diagonal.

Table 4 shows intercorrelations among narrative features, separately for perpetrator and victim stories. For perpetrator stories positive and negative exploration are positively correlated, and positive exploration is associated with both growth and resolution. For victim narratives, positive exploration was associated with negative exploration and growth, but not resolution, and negative exploration was related to both growth and damage conclusions. Finally, cross-event intercorrelations are presented in Table 4. The results indicate that positive and negative exploration are more consistent across event types within individuals' narratives, compared to other narrative features; this is consistent with the idea that these features of narrative share a common underlying exploration tendency.

Table 4. Intercorrelations among the narrative features for perpetrator and victim stories.

	Victim Positive Exploration	Victim Negative Exploration	Victim Growth	Victim Damage	Victim Resolution
Perpetrator Positive Exploration	.31**	.18	-.005	-.07	-.05
Perpetrator Negative Exploration	.16	.61***	.09	.29**	-.22*
Perpetrator Growth	.06	.06	.07	-.03	.01
Perpetrator Damage	-.07	.15	-.07	.07	-.05
Perpetrator Resolution	.02	.15	.05	.01	.20

Note.

^*= p<.05,

^**= p<.01,

^***= p<.001

Preliminary analyses showed there were no effects of order (victim or perpetrator narratives first) on the variables of interest; likewise, there were no effects of gender on those variables. Order and gender are not included in our analyses below and are not further discussed.

Do Perpetrator and Victim Narratives Differ in Exploration, Growth, Damage, and Resolution?

To evaluate differences in how participants narrated perpetrator and victim experiences, we conducted a general linear model with the five narrative features as dependent measures, and event type (perpetrator or victim) as a within-subjects independent variable. There was a significant main effect of event type, F (5, 74) = 4.89, p = .001, η_p² = .25. Univariate tests showed that growth conclusions were more frequent in perpetrator narratives (EMM= .06, SEM= .02) than in victim narratives (EMM= .01, SEM = .01; F (1, 78) = 7.83, p< .01, η_p² = .09), a finding mirrored in other work (Mansfield et al., 2010). The univariate tests also showed that damage conclusions occurred more frequently in victim narratives (EMM = .07, SEM = .01), than in perpetrator narratives (EMM = .01, SEM = .01; F (1, 78)= 13.31, p = .001, η_p² = .15). Perpetrator and victim events did not differ in their levels of positive exploration, negative exploration, or resolution. These findings suggest that both types of events contribute to narrative identity processes in overlapping and also distinct ways – such that narrating perpetrator events provides the possibility of positive conclusions and growth more so than narrating victim events, and narrating victim events poses a risk for drawing damage conclusions about the self. These findings are broadly consistent with previous work, but do not address how differences in narration were connected to differences in ratings of current emotions about the events and in perceptions that the events were identity central.

Does How People Narrate an Event Matter for Emotions and Identity Centrality?

We used HLM 7 (Raudenbush & Bryk, 2002) to analyze associations between narrative features, basic negative emotions, moral emotions, and identity centrality for victim and perpetrator experiences. HLM is the appropriate choice for our analyses because ratings of negative emotions, moral emotions, and identity centrality are nested within people and event type. Using HLM avoids misestimating standard errors when data are inherently dependent, by including in the statistical analyses unique random effects for each individual in the sample (Raudenbush & Bryk, 2002). It also allows one to model heterogeneity of regression which occurs when the relationship between the measured outcomes and predictors vary for individuals being studied, as we expected would be the case in our study.

In addition to modelling main effects of event type and narrative features in predicting identity and emotion ratings, we also included interactions between event type and each narrative feature in our initial models. Main effect terms were group centered prior to analysis, and interaction terms were computed on the centered main effect variables. Because at the single event level, narrative features were sparsely distributed at the upper end of our scales, we also ran all the analyses reported below using bivariate versions of the coded narrative features; results were identical to those reported below.

Basic Negative Emotions

In our first HLM model, we evaluated relationships between narrative features (exploration, growth, damage, and resolution) and event type and basic negative emotions. Only level 1 predictors were included in the models, as follows:

Level-1

Basic Negative Emotions_i person_j = β0j + β1j*(Event Type) + β2j*(Positive Exploration) + β3j*(Negative Exploration) + β4j*(Growth) + β5j*(Damage) + β6j*(Resolution) + β7j*(Positive Exploration*Event Type) + β8j*(Negative Exploration*Event Type) + β9j*(Growth * Event Type) + β10j*(Damage * Event Type) + β11j*(Resolution * Event Type) + rij

The level 1 model is analogous to calculating a separate regression equation for each person in our study based on a “sample” of his/her responses to the 12 events elicited (6 narratives for two types of events). Because story type was dummy coded (perpetrator = -.5; victim = .5), β1j describes how ratings of negative emotions change when participants report perpetrator vs. victim experiences. The β2j through β6j slope coefficients describe how negative emotion ratings change as negative exploration, positive exploration, growth conclusion, damage conclusions, and resolution (respectively) deviate from participants' average amount of those narrative features. The remaining slope coefficients describe how ratings change as a function of interactions between narrative features and the type of story.

Table 5 displays coefficients and p values from this model. Event type was significantly related to degree of negative emotion reported with victim experiences being associated with more negative emotion than perpetrator experiences. Additionally, when participants narrated harm events with more than their typical level of resolution, they also reported decreased negative emotion about the event. Note that no interactions with type of event were found.

Table 5. Predicting Basic and Moral Negative Emotions and Identity Centrality from Story Type and Narrative Features.

Dependent Variable	Independent Variable	Coefficient (SE)	t-Ratio	Approximate d.f.	p-value
Basic Negative Emotions
	Intercept, β0	2.64 (.10)	25.44	83	<.001
	Event Type, β1	.85 (.00)	9.58	829	<.001
	Positive Exploration, β2	.19 (.15)	1.32	829	.19
	Negative Exploration, β3	.06 (.06)	1.00	829	.32
	Growth, β4	.35 (.25)	1.41	829	.16
	Damage Conclusions, β5	.09 (.26)	0.37	829	.72
	Resolution, β6	-.40 (.12)	-3.47	829	<.001
	Positive Exploration by Event Type, β7	.18 (.28)	0.64	829	.52
	Negative Exploration by Event Type, β8	-.11 (.11)	1.00	829	.32
	Growth by Event Type, β9	-.14 (.50)	-0.29	829	.78
	Damage Conclusions by Event Type, β10	-.09 (.58)	-0.15	829	.88
	Resolution by Event Type, β11	-.22 (.26)	-0.85	829	.39
Moral Negative Emotions
	Intercept, β0	2.58 (.10)	26.62	83	<.001
	Event Type, β1	-1.25 (.12)	-10.36	829	<.001
	Positive Exploration, β2	-.05 (.14)	-0.39	829	.70
	Negative Exploration, β3	.13 (.09)	1.43	829	.15
	Growth, β4	.23 (.30)	0.76	829	.45
	Damage Conclusions, β5	-.19 (.32)	-0.57	829	.57
	Resolution, β6	-.23 (.19)	-1.23	829	.22
	Positive Exploration by Event Type, β7	-.36 (.36)	-1.00	829	.32
	Negative Exploration by Event Type, β8	-.06 (.17)	-0.39	829	.70
	Growth by Event Type, β9	-.11 (.65)	-0.17	829	.87
	Damage Conclusions by Event Type, β10	-.06 (.61)	-0.10	829	.93
	Resolution by Event Type, β11	.37 (.35)	-1.04	829	.30
Identity Centrality
	Intercept, β0	3.26 (.13)	24.17	83	<.001
	Event Type, β1	.55 (.10)	5.46	829	<.001
	Positive Exploration, β2	.27 (.15)	1.79	829	.08
	Negative Exploration, β3	.10 (.07)	1.50	829	.14
	Growth, β4	.32 (.30)	1.04	829	.30
	Damage Conclusions, β5	.84 (.21)	3.91	829	<.001
	Resolution, β6	-.44 (.16)	-2.72	829	<.01
	Positive Exploration by Event Type, β7	.22 (.34)	0.63	829	.52
	Negative Exploration by Event Type, β8	-.33 (.16)	-2.07	829	.04
	Growth by Event Type, β9	.31 (.64)	0.49	829	.63
	Damage Conclusions by Event Type, β10	.05 (.53)	0.10	829	.92
	Resolution by Event Type, β11	.06 (.42)	0.15	829	.88

Moral Negative Emotions

Using the same model, we then asked how narrative features (exploration, growth, damage, and resolution) and event type were related to moral negative emotions. We followed the same procedures as above. Table 5 displays coefficients and p values for moral negative emotions. We found that event type was significantly related to degree of moral emotions reported, with perpetrator experiences being associated with a higher degree of moral emotions than victim experiences.

Identity

We then repeated the above model, this time predicting participants' identity centrality ratings. Table 5 displays coefficients and p values from this analysis. Event type was significantly related to centrality of identity, with victim experiences being associated with higher identity centrality than perpetrator experiences. When participants narrated an event with more damage conclusions than was typical for them, they reported that the event was more identity central. When they narrated an event with more resolution than was typical for them, they reported that the event was less identity central. Finally, there was a significant interaction of event type with negative exploration, suggesting that negative exploration had different implications for identity depending on the type of event being narrated. To follow this interaction, we conducted simple slopes tests using the online utilities created by Preacher and colleagues (Preacher, Curran, & Bauer, 2006). Simple slopes tests showed that the effect of negative exploration on identity was significant and positive for perpetrator events, Slope = 0.27(0.11), t=2.34, p=0.02; but near zero and non-significant for victim events, Slope = 0.07(0.11), t=-0.57, p=0.57. This effect is shown in Figure 1; as seen there, victim events were generally higher in identity centrality but when participants engaged in negative exploration, perpetrator events were equally highly rated.

Figure 1. Interaction between Event Type and Negative Exploration on Ratings of Identity Centrality.

Taken together, these findings suggest that perpetrator and victim events differ in their emotional and self-regulatory challenges. Perpetrator events were associated with higher guilt/shame but lower levels of basic negative emotions and lower ratings of identity centrality; these results mirror those reported above with the general linear model. However, the way participants narrated the experiences also mattered, particularly for basic negative emotions and for identity centrality. Resolution was associated with lower basic negative emotion and less identity centrality. Damage conclusions were associated with higher identity centrality. Moreover, exploration of the negative implications of an event yielded increased perceptions of identity centrality, but only for perpetrator events.

Relationships between narrative features, identity and emotion ratings, and well-being

Our last question was whether and how narrative features, identity, and emotions regarding events might be related to broader indices of well-being. For this analysis, we aggregated across events to create average scores for the narrative features and the identity and emotion ratings, and computed correlations with a composite measure of well-being; these correlations are displayed in Table 6. Aggregate identity and emotion variables were correlated with overall well-being. However, aggregate narrative measures showed no significant correlations with the overall well-being measure. Note that aggregate measures of emotion and identity centrality were quite highly correlated across victim and perpetrator events (r's between .40 and .78, p's < .01). Further, aggregate narrative features were correlated with the aggregate emotion and identity measures, especially for perpetrator events.

Table 6. Intercorrelations among narrative feature means, well-being, moral negative emotions and basic negative emotions in perpetrator and victim events.

	Well-Being; all events	Well-Being; identity centrality > 3	Identity Centrality	Moral Negative Emotions	Basic Negative Emotions
Perpetrator Well-Being	--	--	-.22*	-.27*	-.34**
Perpetrator Positive Exploration Mean	.09	.31	.16	-.12	-.04
Perpetrator Negative Exploration Mean	-.02	-.09	.26*	.20	.15
Perpetrator Growth Conclusion Mean	.15	.21	-.06	-.04	-.07
Perpetrator Damage Conclusion Mean	-.19	-.05	.32**	.37***	.34***
Perpetrator Resolution Mean	.20	-.14	-.20	-.24*	-.32**
Victim Well-Being	--	--	-.23	-.09	-.47***
Victim Positive Exploration Mean	.12	-.04	-.03	.01	-.19
Victim Negative Exploration Mean	.08	-.08	.26*	-.07	-.11
Victim Growth Conclusion Mean	.15	.23	.06	.17	-.13
Victim Damage Conclusion Mean	-.03	-.21	.05	-.02	.07
Victim Resolution Mean	-.12	.10	-.13	-.12	-.10

Note.

^*= p<.05,

^**= p<.01,

^***= p<.001

The absence of associations with aggregate narrative features and well-being is not consistent with previous research, where growth, exploration, and resolution have been linked to well-being (e.g., Bauer et al., 2010; Lilgendahl & McAdams, 2011). One possibility is that such relationships are primarily evident for events deemed identity relevant, and therefore the relations in our more heterogeneous sample of events are attenuated. To explore this possibility, we recomputed correlations between aggregate narrative features and well-being, but this time for only events rated 3 or higher on the identity centrality measures; these correlations are also presented in Table 6. Note that participants' events were reported as relatively average to below average in centrality but that each person reported up to 12 events; thus, the sample size for the analyses reported in Table 6 varied from 66 (victim) to 77 (transgression) participants. As can be seen, when looking only at events rated at least somewhat central to identity, our findings suggest trend-level relationships between growth in the narratives, and higher well-being, consistent with previous work.

Taken together, these data suggest that the path from narrating experiences of harm to overall well-being involves both the type of event and the way the event is narrated. Moreover, that path may operate indirectly via the effects of narration on emotions and perceptions of events as identity central.

Discussion

This study aimed at exploring how narration might resolve emotion and identity challenges posed by experiences of interpersonal harm. Our findings suggest that people narrate harm events differently depending on whether those harm experiences were from the victim or perpetrator perspective. Narratives about perpetrator experiences entail more frequent growth conclusions, while those about victim experiences entail more frequent damage conclusions. However, victim and perpetrator experiences were similar in terms of exploration and resolution. These findings are consistent with previous work on perpetrator and victim experiences (Baumeister et al., 1990; Wainryb et al., 2005), while extending those findings to narrative features with known relevance to identity and self. Specifically, the findings suggest that narrating perpetrator and victim experiences offers different possibilities for meaning-making, with victim experiences presenting more dangers in terms of construing the self as damaged, and perpetrator experiences providing more opportunities for constructing a sense of growth. Notably, it is not a question of these events affording “more or less” meaning, but rather, posing distinctive opportunities and risks (see Lilgendahl et al., 2013; Pasupathi & Wainryb, 2010; Recchia, Wainryb, Bourne, & Pasupathi, in press; Wainryb & Pasupathi, 2010).

In line with this thinking, our findings also suggested that the type of experience (victim and perpetrator) mattered for emotion and identity centrality. Victim experiences evoked more intense basic negative emotion and were deemed more identity central, while perpetrator experiences evoked more intense moral emotions. These findings are consistent with other work on traumas and transgressions (Mansfield et al., 2010), and with established differences in the way perpetrator and victim experiences are narrated (Baumeister et al., 1990; Schütz, 1999; Wainryb et al., 2005). Such differences may also be related to self-protective ways of construing one's own hurtful actions as less connected to the self (Baumeister et al., 1990; Pasupathi & Wainryb, 2010; Wainryb et al., 2005).

Beyond the type of experiences, however, the ways in which experiences were narrated also mattered, sometimes in simple, and sometimes in complex ways. In considering narrative features, exploration (positive and negative) reflects a kind of processing of experiences, while growth, damage, and resolution reflect potential outcomes of that processing. Generally, our data suggested that resolution was related to emotion and identity outcomes. Across victim and perpetrator experiences, narratives with greater resolution were associated with reduced basic negative emotion and reduced identity centrality. This is consistent with experimental work in our own laboratory (Pasupathi et al., under review), in which closure in narratives were linked to steeper declines in sadness and anger. It is also consistent with work showing that distancing reappraisals may render events less emotionally intense when recalled (Kross & Ayduk, 2008). Growth and damage had fewer relationships to emotion and identity; with damage being linked positively to identity centrality. This finding suggests that while narrating growth may benefit people's functioning in many ways, it does not necessarily always do so via reducing distress or altering perceptions of the identity-centrality of events, at least for more everyday events.

In fact, as we have articulated in other work (Pasupathi & Wainryb, 2010; Pasupathi et al., under review), narrating can also create and maintain difficulties. Narratives with damage conclusions were associated with higher identity centrality. Such conclusions were also more likely in the context of victim experiences. One risk arising from the combination of victim experiences and narrative processes, then, is the construction of an identity as a victim – when being the target of harm becomes central to one's sense of self. Feeling that experiences of victimization are central to one's identity may be problematic in that this is likely to have implications for well-being (e.g., Berntsen & Rubin, 2007; Park, Zlateva, & Blank, 2009), and could be conceptually linked to learned helplessness and reduced self-efficacy (Abramson, Seligman, & Teasdale, 1978; Albert Bandura, 1977); these ideas warrant further direct testing.

Although our data allow testing of interactions between narrative features and event type in relation to emotion and identity outcomes, we found only one case where the effect of a narrative feature depended on the type of event narrated. For identity centrality, the implications of negative exploration depended on whether participants narrated perpetrator or victim events, with identity centrality significantly and positively related to negative exploration only for perpetrator events. Because perpetrator events, by definition, entail more agentic involvement than do victim events, one possibility is that exploration of experiences is more likely to result in identity centrality when the exploration concerns one's own agency; by contrast, for events with less agentic involvement, the identity implications for the experience may be contained in the damage conclusions. However, the findings overall suggest that the narrative features we examined were generally operating similarly regardless of the type of story. Although this may suggest that growth, damage, resolution, and exploration operate similarly across different types of experiences, it could be that other narrative features might not operate similarly. For example, indications of perspective taking (e.g., Wainryb et al., 2005), or the articulation of more specific types of threats or challenges to identity and emotion (Mansfield, in preparation), may be more likely to operate differently for different types of events.

Finally, our findings suggest that reporting strong negative emotions or high identity centrality in relation to interpersonal harm experiences is associated with lower levels of well-being. Aggregate narrative measures, by contrast, were not linked to well-being directly. This contradicts some prior work, and there may be at least two sets of reasons why that could be the case. First, the elicitation of 6 victim and 6 perpetrator events necessarily results in the narration of events of varying importance. Other studies have focused on single events in each category or on events already deemed identity central in some way, so that participants probably reported on the most memorable and important event, increasing the likelihood of direct relationships between narration and well-being. Previous research suggests that narrative features have stronger relationships to well-being when an experience is clearly identity central (Baerger & McAdams, 1999; Waters & Fivush, 2014). Looking at heterogeneous events may reduce the magnitude of relationships between narrative features and indices of adaptive functioning because less identity central events, regardless of how they are narrated, do not bear on well-being (Baerger & McAdams, 1999; Waters & Fivush, 2014). Indeed, a look at correlations in our own data for only the more identity-related events suggests that this might be the case within our sample as well – when we restricted our examination to only events with stronger identity centrality, the magnitude of correlations increased (doing so, however, also necessarily compromised our power slightly.)

A second way to think about the findings is that they are consistent with theoretical perspectives on narrative and self-development outlined elsewhere (McLean et al., 2007; Pasupathi, 2001; Weeks & Pasupathi, 2011). Repeated everyday narration may contribute to broader indices of self and well-being indirectly, through an emerging sense of the self that is based on many everyday events. Thus, the way a participant narrates a specific, everyday perpetrator or victim event may not be the only critical factor in maintaining well-being. Narrative construction of interpersonal harm may affect well-being indirectly, via the implications of narrative construction for down-regulating negative emotions and for drawing conclusions about the identity implications of an experience. This interpretation is consistent with theoretical models in the literature that suggest everyday narration of experiences has indirect effects on well-being through the implications of such narration for self and identity beliefs (McLean et al., 2007). Specifically for our findings, narrative construction of the self as damaged, in relation to experiences of harm, is linked to viewing harm events as central to identity; which is in turn related to lower well-being.

In fact, there is an apparent contradiction in the literature on trauma and identity centrality that may be addressed by the present findings. The contradiction stems from two sets of research findings. Articulating growth meanings in narratives about important and challenging life events, including traumatic ones, is consistently related to higher well-being (Lilgendahl & McAdams, 2011; McLean et al., 2007; Reese et al., 2011) as well as to more adaptive emotion regulation strategies (Cox & McAdams, 2014). But, people who rate traumatic events as identity central report lower well-being (Berntsen & Rubin, 2007). Growth conclusions seem conceptually related to identity centrality – that is, if people view an experience as having helped them to grow, one might expect that they should also view the experience as more identity central. The present findings suggest that this apparent contradiction arises because growth narration and identity centrality are not always positively linked. Rather, identity centrality of negative events is related to construing the events as having damaged the self.

Narratives that construct the possibility of growth result in the distancing of a present or future self from the negative experience that is being narrated. In other words, if I feel I have grown from an event, I no longer am the person who engaged in or suffered from that harmful act. I'm a “new”, different person. Thus, a growth conclusion could relegate an event to being less identity central because the experience is construed as linked to a past and different self (Ross & Wilson, 2003; Wilson & Ross, 2003). This reasoning suggests that the implications of growth conclusions for perceptions that an event is identity central may be quite variable, consistent with our finding that growth conclusions in narratives were unrelated to identity centrality.

As with any study, this one has some important limitations. First, we used a student sample, and students may be limited in the nature of their experiences. In this case, however, many other studies of victim and perpetrator narration (Wainryb et al., 2005), as well as of narration and well-being outcomes (McAdams et al., 2001; Smyth & Pennebaker, 2008) suggest that some generalizability is likely, if not proven. Other work, though, suggests that older adults may be less distressed about events like these than are younger adults (Pasupathi & Mansour, 2006; Lilgendahl, McLean, & Mansfield, 2013), but that middle-aged and possibly older adults may benefit more from narrating growth (Lilgendahl & McAdams, 2011). Second, we had a single time-frame for assessment, thus, all of our findings are correlational in nature. Again, other work suggests that narration has prospective implications for identity, emotion, and well-being (Lilgendahl & McAdams, 2011; Mansfield et al., under review; Pasupathi et al., under review; Reese et al., 2011). However, theoretical frameworks suggest bidirectional and reciprocal relationships over time between narrative, identity, emotion, and well-being. So, more well-being may relate to more adaptive narration, which in turn maintains or enhances well-being over time (Mansfield et al., in press), or greater identity centrality of an event may be connected to distinct ways of narrating the event, with distinct implications for well-being. Narratives themselves may change over time in ways that move towards (or away from) adaptive content. Third, we focused on a particular approach to measuring identity centrality, namely, participants' ratings of centrality. It is worth noting that other approaches to centrality, such as nominating events that are part of one's life story, might yield different findings. These are clearly issues for future work.

More broadly, and despite these limitations, our findings suggest that distinctive negative experiences may offer distinct challenges and opportunities for self and identity development. They re-affirm the importance of how experiences are narrated as one process by which experiences shape emotion, identity, and well-being. They also suggest that some of those relationships are indirect – in that we may become the stories we tell, in part via the emotional and identity conclusions those stories create.

Highlights.

• Victim and perpetrator events provide distinct challenges to identity and emotion.

• Both event type and narration are related to identity and emotion outcomes.

• Everyday narration may be linked to well-being indirectly via emotion and identity.