Storied Lessons: Learning from Anger in Childhood by Narrating

Abstract

We evaluated whether narrating anger-provoking events promoted learning from those events, as compared with other responses to anger, and whether the effectiveness of narrative depended on age. In addition, we tested relations between anger-reduction and learning and in a subset of participants, between narrative quality and learning. 248 youth (8 to 17 years old) recalled an anger-provoking experience, and were randomly assigned to one of four activities: recalling the event a second time, narrating the event, and distraction (via video game play or conversation). Youth then recalled the event one last time, and rated the extent to which they had learned from that event. Younger children reported more learning when they had narrated their experience. Older youth reported more learning when they had narrated the event more frequently prior to participation. Stronger reductions in anger following regulation were associated with greater self-reported learning. Finally, more elaborative and less resolved narratives were associated with greater self-reported learning.

Experiences of interpersonal anger are replete with the potential for insights about self and others. One route by which people turn experiences of anger into expanded understandings of their interpersonal worlds is via reflecting on those experiences, often by narrating them. A broad literature on children’s and adolescents’ narration of emotional experiences with parents suggests that parents use the co-narration of emotional events to scaffold youth’s understandings of emotion, complex social and moral experiences, and emerging sense of self and identity (Bird & Reese, 2006; Cleveland, Reese, & Grolnick, 2007; McLean & Mansfield, 2012; Recchia, Wainryb, Bourne, & Pasupathi, 2014; Reese, Yan, Jack, & Hayne, 2010). Parents do so by helping children to consider elements of their experiences – early in childhood focusing on helping children to tell a complete story (e.g., Reese, Haden, & Fivush, 1993), and later focusing more on the psychological elements of events – beliefs, desires, and emotions – that make events make sense (e.g., Fivush & Kuebli, 1998; Fivush, Berlin, Sales, Mennuti-Washburn, & Cassidy, 2003; Fivush, Bohanek, Robertson, & Duke, 2004; Recchia, Wainryb, Bourne, & Pasupathi, 2014; Reese, 2002). Moreover, parents do this in ways that are responsive to youth’s developing social and cognitive capacities, and in ways that are tuned to specific emotions. For example, parent-child narration about anger focuses on the self more than parent-child narration about sadness or other emotions (Fivush et al., 2003).

Variations in the extent to which parents encourage elaborative, meaning-focused narration of emotional events are associated longitudinally with inter-individual differences in youth’s emotion regulation and self-understanding (Bird & Reese, 2006; Farrant & Reese, 2000; Fivush, Haden, & Reese, 2006). As youth get older, they become increasingly capable of constructing elaborated narratives in social and solitary contexts with less intensive support and scaffolding from others (de Silveira & Habermas, 2011; Habermas & Bluck, 2000; Habermas,

Negele, & Meyer, 2010; McLean & Breen, 2009; McLean & Mansfield, 2012; Pasupathi & Wainryb, 2010a). In sum, narrating emotional experiences in general, and anger events in particular, can promote youth’s learning about themselves and their social worlds. However, existing work on the learning potential of narrative is incomplete.

One major concern is that extant work has not compared narration to other responses to emotions. Narration offers unique possibilities for learning about oneself and others in relation to emotional events because narration requires articulating information about oneself and others’ psychological experience (McLean & Breen, 2009; McLean & Mansfield, 2012; Pasupathi & Wainryb, 2010a; Recchia et al., 2014). However, youth may not always narrate anger-provoking experiences, and it is possible that learning from such experiences could occur simply by recalling the event privately. Further, in managing anger, people often seek to reduce the intensity of their anger, which may help control responses and avoid lasting interpersonal conflict.

Strategies that reduce intensity, however, may result in the loss of learning opportunities. Distraction - or shifting attention away from a distressing event – is a strategy available to people of all ages, from infancy to late life (Amirkhan & Auyeung, 2007; Rusting & Nolen-Hoeksema, 1998; von Salisch & Vogelgesang, 2005; Zimmermann & Iwanski, 2014). It may take various forms, from simple attention shifts towards neutral material (Scheibe, Sheppes, & Staudinger, 2015; Zimmer-Gembeck & Skinner, 2011) to actively pursuing immersive experiences such as socializing or video gaming. Distraction is preferred by children (Eisenberg, Fabes, Nyman, Bernzweig, & Pinuelas, 1994; Kopp, 1989) perhaps because it is highly effective at reducing distress in the short-term (Suls & Fletcher, 1985; Pasupathi, Wainryb, Mansfield, & Bourne, 2017, Wainryb, Pasupathi, Bourne, & Oldroyd, 2018). This effectiveness is short-lived, and requires that people continue to distract and avoid engaging with a distressing event (Kross & Ayduk, 2008; Wainryb et al., 2018; Pasupathi et al., 2017). It also seems obvious that, relative to the exploration and elaboration afforded by narrating an experience, distraction would result in less learning. However, the reality is more complex.

Although emotion productively focuses attention on important aspects of the environment (Parrott, 1991; von Salisch & Saarni, 2011), emotion can disrupt cognition (e.g., Drevets & Raichle, 1998; Davis & Levine, 2013; Davis, 2016). To the extent that emotion is cognitively disrupting, reducing anger might facilitate greater learning regardless of strategy. There is some evidence that when youth (ages 6 to 13 years) effectively reduce their emotional distress in response to an emotion elicitation, they are better able to learn from subsequently presented, emotionally neutral, and unrelated material (Davis & Levine, 2013), presumably because their attentional focus is broadened (Davis, 2016; Fredrickson & Branigan, 2005). In Davis and colleagues’ work, youth were randomly assigned to reappraise or ruminate about a sad film, and then were asked to watch a neutral educational film. Reappraisal was more effective at reducing sadness, and was associated with more recall of the educational film and a broader focus of attention on that film. So, it is possible that if distraction reduces distress temporarily, youth may be more able to learn from an anger event on a subsequent exposure to that event.

Compared to distraction, narration offers reductions of emotional distress that are less strong, but which may be longer-lasting (Pasupathi et al., 2017; Wainryb et al., 2018). Thus, both narration and distraction might promote learning via distress reduction.

For Learning, Some Stories may be Better than Others

Not all narratives are created equal. In narrating emotional experiences, people vary in the extent to which their narratives elaborate on what happened and the psychological underpinnings and implications of the event (Fivush et al., 2006; Köber, Schmiedek, & Habermas, 2015; Lilgendahl & Mcadams, 2011; McLean & Breen, 2009; Pasupathi, Mansour, & Brubaker, 2007; Pasupathi & Wainryb, 2010b). People also vary in the extent to which their narratives resolve experiences or draw positive conclusions about the events and the self (e.g., Adler & Olin, 2012; Bauer, McAdams, & Pals, 2008; Bohanek, Fivush, & Walker, 2005; Lilgendahl & McAdams, 2011). Some of these variations in narrative quality are linked to variations in the effectiveness of narration for emotion regulation (Kross & Ayduk, 2008; Pasupathi et al., 2017; Pennebaker & Seagal, 1999).

Narratives that elaborate on experiences and their potential meanings are more likely to promote learning about oneself and other people. In fact, such elaboration can be understood as prima-facie evidence of that learning. Further, given that elaboration and positive resolution promote larger reductions in anger – these features may also have indirect benefits for learning. These same features of narrative show developmental changes across childhood and adolescence.

Developmental Changes in Narrative Capacities and Emotion Regulation Preferences

Children begin to narrate their experiences very early (Fivush et al., 2004; Haden, Haine, & Fivush, 1997), but their ability to create coherent, complete, and complex narratives continues to develop until early or even mid-adulthood (Habermas & Bluck, 2000; Köber et al., 2015; Pasupathi & Wainryb, 2010b). This is especially the case for elaboration and exploration in narrative, which may continue to increase in sophistication until middle adulthood (Köber et al., 2015; Mclean & Breen, 2009; Reese, Yan, Jack, & Hayne, 2010). Further, while youth of all ages can engage in and may prefer distraction (von Salisch & Vogelgesang, 2005; Zimmer-Gembeck & Skinner, 2011; Zimmermann & Iwanski, 2014), narration presents more developmental complexities. As noted, very young children engage in narration with their parents (Fivush et al., 2004), but the frequency of narrating experiences of anger increases as children become adolescents (Shipman, Zeman, Nesin, & Fitzgerald, 2003). Likewise, adolescents also report greater use of cognitive strategies for emotion regulation that are related to narration, such as reappraisal (Band & Weisz, 1988; Waters & Thompson, 2014; Zimmermann & Iwanski, 2014).

Taken together, these developmental shifts suggest that older youth may benefit more from narration for learning, relative to younger children. Because adolescents are capable of generating more exploratory and sophisticated narratives, and because cognitive approaches to emotional experiences may be more comfortable for them, they are likely to “get more” out of narrating their experiences.

Summary and Present Study

The present study examined how narration might promote learning from anger experiences. We drew on a study of narrative and anger regulation conducted in our laboratory (Wainryb et al., 2018). One prior publication from this dataset reported on reductions in youth’s anger as a function of narrating their experience (Wainryb et al., 2018). That publication compared narrative to other approaches to anger regulation. The present publication is distinct in focusing on analyses of narrative and learning, which involve different literatures, analyses, and hypotheses as well as a different outcome variable.

In the present study, we posed four interrelated questions about narrative and learning from anger-provoking events. First, we evaluated whether children and adolescents who narrated angry experiences benefit more in terms of perceived learning, relative to youth who distract themselves from the angry experience, and relative to a control condition. We hypothesized that narrating would promote learning better than distraction, and that distraction might promote more learning than a remembering control condition due to reduced anger. Second, we tested whether the impact of narration on learning was different depending on age, across an age range associated with substantial changes in narrative capacities (8 to 17 years). We hypothesized that narrative might be more effective for learning among adolescents, compared to children, based on their greater narrative capacities. Third, we examined whether reduced anger was associated with more reported learning. Fourth, in a subset of participants, we examined relationships between narrative quality and self-reported learning. We hypothesized that aspects of narratives related to elaboration and exploration of meaning, including perspective-taking, would be associated with more self-reported learning.

Methods

Participants

Two hundred and forty eight youths aged 8 to 17 were recruited from community locations in a Rocky Mountain metropolitan area for a study on narrative and emotion regulation (Wainryb et al., 2018). The sample consisted of 121 males and 127 females (Mage=12.55, SD=2.89; evenly distributed across the full age range. Eighty four percent of the sample identified as European American, 7% as Latino/a, 4% as African American, 3% as Asian American, 1.6% as Native American, 1.6% as Pacific Islander, with the remaining 1% choosing to not respond. Youth were paid $25.

Procedure

Participants were randomly assigned to one of four regulation conditions: narration, game-play, conversation, or control. We oversampled the narrative condition to provide power for analyzing relations between narrative quality and learning. Following consent and assent, youth were trained to use the iPad employed for data collection. Those participants assigned to the game play condition were also taught how to play the matching game “AniMatch”. After a brief rest period (2 minutes), participants completed their first emotion rating (Baseline).

Initial exposure.

Following the first emotion rating, the autobiographical memory was elicited in two steps (1) Reminding youth that everyone feels mad sometimes, thus normalizing experiences of anger, and (2) Asking participants to remember a time in their own life when “someone did or said something and they felt really mad at that person”. Once the participant indicated that they had thought of an experience, they were instructed to, “spend a few minutes thinking about that time when…you felt really mad”. Participants were then left alone for two minutes, after which time they completed another emotion rating (Exposure).

Regulation.

Following exposure, participants were instructed to engage in one of four regulation activities. In the narrate condition, participants were asked to tell a trained research assistant about their angry memory. They were prompted to “Tell me everything that happened, how you felt about it, and also what you learned from it.” Participants in the conversation condition were asked to tell the research assistant about a time when, “someone did or said something and you felt really happy with that person” and given the same prompting. Participants in the game play condition played a matching game on the iPad for two minutes. Because preliminary analyses indicated that game play and positive narration did not differ in terms of reported learning, we aggregated these two conditions into a single “distract” condition. The control group was instructed to think about their previously nominated angry memory for an additional two minutes. Following regulation, all participants completed a third emotion rating (Regulation).

Re-exposure:

Regardless of condition, following regulation, immediate re-exposure to the memory was prompted as follows, “For the last time today we are going to have you remember again, that same time when you felt really mad”. Participants were left alone for two minutes and were then instructed to complete their fourth and final emotion rating (Re-exposure).

Conclusion of Session:

Participants reported on how long ago this memory had occurred, how important the event was to them, how often they had told the story previously, and how much they learned from it.

Measures

Learning.

Participants rated the extent to which they had learned from the event about themselves and about others on 5-point Likert-type scales with 1 indicating “not at all” and 5 indicating “very much”. These ratings were highly correlated, r(247) = .51, p < .001, and we aggregated them into an overall learning variable for analysis (α = .67).

Age of Event.

We asked participants when the event occurred (this past week, this past month, many months ago, more than a year ago, and more than 3 years ago). Fifty-seven percent of participants reported on events that occurred within the past year, with the remainder focusing on older events.

Importance of event.

Participants rated “how important is this event to you now” on a 5-point Likert-type scale where 1 indicated not important and 5 indicated very important; M(SD) = 2.7(.12) for the entire sample.

Prior narration of event.

Participants rated the extent to which they previously told others about the event on a 5-point Likert-type scale where 1 indicated no prior telling and 5 a great deal of previous narration; M(SD) = 2.7(1.3) for the entire sample.

Anger.

Participants rated their anger after initially recalling the angry memory, again after regulating their emotions, and again after recalling the memory on a 5-point Likert-type scale where 1 indicated no anger at all and 5 a great deal of anger.

Narrative Quality.

For participants in the narrate condition, narratives were transcribed and coded for six narrative features: elaboration, resolution, positive meaning-making, and perspective-taking. We employed established schemes to code elaboration, positive meaning-making, and resolution (Pals, 2006; Pasupathi, Billitteri, Mansfield, Wainryb, Hanley, & Taheri, 2015). We developed a simple scheme to capture perspective-taking for this project, drawing on traditions in moral development (Wainryb, Brehl, & Matwin, 2005). To establish initial reliability, research assistants were trained on the narrative features and definitions of the scales used for coding (see below), after which they independently coded an initial set of narratives to a criterion high agreement level. Coders then independently coded the narratives with an additional 25% of the narratives double-coded for reliability.

Factual Elaboration.

Factual elaboration captures elaboration of information that can be objectively verified, such as perceivable actions (e.g. He was crying). Narratives were given a score from 0, indicating an absence of factual information to 3, indicating elaborated factual information. ICC=.93.

Interpretative Elaboration.

Interpretive elaboration emphasizes the subjective, meaning-laden, and internal aspects of experiences. Examples of this category include statements describing emotions, evaluations, goals, thought and personal judgments. Narratives scored from 0, indicating an absence of interpretive information to 3 indicating a high level of interpretive information. ICC=.86

Exploratory Processing.

Exploratory processing captures the extent to which the participant tries to forge a connection between past events and current identity. Narrative qualities indicative of exploratory processing include open and honest description of emotion; description of exploratory processes (e.g. I thought about this for a while); acknowledgement of past or present uncertainty (e.g. I’m not quite sure about this); contextual considerations (e.g. I was sick and hadn’t slept very well the night before); and statements of values and beliefs (e.g. I don’t think reacting with anger is the right thing to do). Narratives were scored from 0 (No exploratory processing, the event is dismissed or minimized) to 4 (High exploratory processing; several different indicators of exploration are present and well elaborated). ICC = .77.

Perspective Taking:

Perspective taking represents the extent that another person’s point of view is present in a narrative. Narratives were scored from 1, indicating that the other’s voice is absent or summarily dismissed (e.g., “I got really mad at my dad because he wouldn’t let me go and that was just stupid”), to 3, indicating that the other’s voice is well represented and viewed as legitimate, though the narrator may disagree (e.g., “And now, the lady who yelled at me before, felt bad”). ICC = .90.

Resolution/Closure:

This dimension captures the extent to which the narrative conveys a sense of emotional resolution. A 1 indicated a narrative low in resolution (e.g., It still haunts me to this day) and a 3 indicated that narrative was high on resolution (e.g. We are best friends again now). ICC=.85.

Positive Meaning Making:

Positive meaning-making involves drawing positive conclusions about the experience, or affirming self-improvement. Narratives were scored from 1, indicating that participant presents no meaning or negative meaning for the experience, to 3 indicating the presence of positive meaning (e.g., I learned that I give good advice; I’m a kinder person than I thought I was; I stand up for what I believe in). ICC = .77.

Results

Preliminary Analyses

For preliminary analyses, we examined whether the events on which youth focused differed as a function of experimental condition, focusing on initial reports of emotional distress on recalling the chosen events, and on ratings of the age of the event, its importance, and prior narration. Table 1 presents descriptive statistics and Table 2 presents intercorrelations for all variables used in the analyses reported below.

Table 1:

Means (standard deviations) for all variables by condition

	Variables	Control (n = 43)	Distract (n = 85)	Narrate (n = 120)
1	Learning	2.97 (1.20)	2.84 (1.19)	2.97 (1.01)
2	Age of Event	2.95 (1.38)	3.13 (1.43)	3.36 (1.40)
3	Importance of Event	2.81 (1.16)	2.71 (1.19)	2.58 (1.19)
4	Prior Narration	2.74 (1.31)	2.69 (1.41)	2.76 (1.33)
5	Anger at Exposure	2.74 (1.05)	2.41 (0.98)	2.45 (0.95)
6	Anger at Re-exposure	2.91 (1.07)	2.36 (1.07)	2.25 (1.08)
7	Aggregated Elaboration			1.76 (0.78)
8	Aggregated Positive Resolution			1.60 (0.48)
9	Factual Elaboration			2.09 (0.75)
10	Interpretive Elaboration			1.65 (0.99)
11	Exploratory Processing			1.54 (1.11)
12	Perspective-Taking			1.33 (0.55)
13	Resolution/Closure			1.98 (0.63)
14	Positive Meaning-Making			1.23 (0.53)

Note: For rows 1–6, variables ranged from 1 (low) to 5 (high). Aggregated elaboration ranged from .33 to 3.33, and aggregated positive resolution ranged from 1 (low) to 3 (high). Factual and interpretive elaboration ranged from 0 (none) to 4 (high). Exploratory processing ranged from 0 to 3 (high). Perspective-taking, resolution/closure, and positive meaning-making ranged from 1 (low) to 3 (high).

Table 2:

Intercorrelations between Primary Variables (including both Aggregated and Coded Narrative Features)

Variables	1	2	3	4	5	8	9	10	11	12	13
1. Learning	‒
2. Age of Event	.07	‒
3. Importance of Event	.26**	−.13**	‒
4. Prior Narration	.41**	.05	.13*	‒
5. Initial Anger	.37**	−.00	.21**	.09	‒
6. Aggregated Elaboration	.41**	.12	.18	.17	.13	--	--	--	--	--	--
7. Aggregated Positive Resolution	−.10	.14	−.07	.04	−.16	--	--	--	--	--	--
8. Factual Elaboration	.16	.04	.11	−.02	.06	--
9. Interpretive Elaboration	.44**	.07	.15	.20*	.16	.37**	‒
10. Exploratory Processing	.36**	.16	.16	.19*	.09	.38**	.71**	‒
11. Perspective Taking	−.05	.05	.14	.13	.04	.09	.18*	.23*	‒
12. Resolution/Closure	−.17	.15	−.14	−.03	−.23*	−.03	.04	.03	.10	‒
13. Positive Meaning Making	.01	.07	.04	.10	−.04	.10	.28**	.36**	.11	.38**	‒

Note: n = 248 for variables 1–5 and 120 for variables 6–13; Columns for variables 6 and 7 are omitted as these variables are aggregates of variables 8–13;

^*p <.05

^**p <.01

Two general linear model (GLM) analyses of these variables as a function of original experimental condition, gender, and age were conducted: one of anger reported on initial recall of the event, and one of ratings of prior narration, importance of event, and age of the event. The results of the GLM of anger at exposure showed that despite random assignment to condition, youth reported different levels of anger at initial exposure to their event depending on condition and prior to engaging in a regulation strategy, F(3,248) = 3.9, p = .009, partial η2 = 05, but there were no age or gender differences in the intensity of anger at initial recall, F’s < 1.0, p’s > .30. Pairwise comparisons showed that participants in the distract conversation condition reported significantly lower initial anger, M(SEM) = 2.1 (.15) than participants assigned to the control condition, M(SEM) = 2.8 (.15) or the distract game condition, M(SEM) = 2.7 (.14). Participants assigned to the narrate condition fell in between, M(SEM) = 2.4 (.09).

The second GLM revealed only effects of age, F(3,235) =7.4, p < .001, partial η2 = .09 and gender, F(3,235) = 6.6, p = .001 partial η2 = .08. Follow-up analyses suggested that age effects were evident for the age of the event, F(1,237) = 6.4, p < .01, partial η2 = .03, and for prior narration, F(1,237) = 16.5, p < .001, partial η2 = .07 and gender effects were evident only for prior narration, F(1,237) = 16.0, p < .001, partial η2 = .06. Girls reported more prior narration, M(SEM) = 3.07 (.13) than did boys, M(SEM) = 2.4 (.13), and older youth and adolescents talked about older events, r(249) = .16, p = .012, and reported more prior narration, r(249) = .25, p < .001 than did younger participants.

Based on these analyses, to the extent that anger on recalling the event initially (exposure), the age of the event, and prior narration are related to learning, these variables needed to be included in testing the effects of narration (versus distraction and control) on reported learning. Prior narrating, r(242) = .40, p < .001, and anger at initial exposure, r(242) = .36, p < .001, were positively and significantly related to reported learning; the age of the event, r(252) = .07, p > .25 was not. So, in testing the effects of narration on learning, compared to the impact of distraction or a control condition, we include emotional distress at exposure, and in evaluating whether any effects of narration on learning are moderated by age or gender, we included prior narration as a control variable.

Question 1: Does narration promote more learning than distraction or mere reexposure?

To address this question, we employed a GLM with aggregated ratings of learning as the outcome measure, and tested the main effects of condition (reexposure-control, distraction combined across conversation and game play, and narration), age, gender, anger-at-recall, and prior narration, as well as two-way interaction terms involving condition, age, and gender. All continuous variables were left continuous in this model. The results of this model show significant main effects of age, F(1,228) = 10.0, p = .002, η2 = .042, gender, F(1,228) = 5.6, p = .019, η2 =.024, and initial anger, F(1,228) = 15.4, p < .001, η2 = .063. In addition, an interaction of condition and age, F(2,228) = 5.2, p = .006, η2 = .043, an interaction of age with prior narration, F(1,228) = 6.88, p = .014, η2 = .026, and an interaction of age with initial anger, F(1,228) = 6.15, p = .014, η2 = .026.

The effect of age was straightforward – across childhood and adolescence, older youth reported greater learning, partial r(242) = .45, p < .001. Likewise, the effect of initial anger was also straightforward, with more intense initial anger on recalling the event associated with greater reported learning about self, partial r(242) = .38, p < .001. Girls reported slightly more learning overall, EMM(SEM) = 2.91 (.09), than did boys, EMM(SEM) = 2.86 (.09).

To clarify the nature of the interactions, we reran the analysis as a regression and probed the interactions using online utilities to calculate simple slopes (Preacher, Curran, & Bauer, 2006). We centered continuous predictors and created two dummy-coded variables to represent the three conditions, with the control condition as the comparison. For the age x condition interaction, we examined perceptions of learning for younger (mean - 1 SD) and older (mean + 1 SD) participants who were assigned to the various regulation conditions (narrate, distract, reexposure control). Younger participants who narrated reported learning more than those in the control condition, B = 0.53, t = 2.49, p = 0.013. By contrast, for older participants who narrated, there was a non-significant decline in reported learning, B = −0.43, t =−1.90, p = 0.059 (see Figure 1).

Figure 1.

Simples slopes of the age by condition interaction: Reported learning for younger and older participants who engaged in narration versus the reexposure control task

Note. * p-value of slope < .05; n = 43 youth 1SD below mean age, n = 52 youth 1SD above mean age.

Decomposing the effect also revealed that age mattered for the impact of distraction on reported learning. Young participants who engaged in distraction reported learning more than participants in the reexposure control condition, B = 0.44, t = 2.02, p = 0.045. The opposite pattern was found for older participants; older youth who engaged in distraction reported learning less than those in the control condition, B = −0.48, t = −2.07, p = 0.040 (see Figure 2).

Figure 2.

Simples slopes of the age by condition interaction: Reported learning for younger and older participants who engaged in distraction versus the reexposure control

Note. * p-value of slope < .05; n = 43 youth 1SD below mean age, n = 52 youth 1SD above mean age.

For the age by prior narration interaction, we examined reported learning for younger (mean – 1 SD) and older (mean + 1 SD) participants who engaged in less (mean - 1 SD) and more (mean + 1 SD) prior narration. The reference group for the dummy-coded condition variables was narrate. Younger participants’ prior narration was not associated with reported learning, B = .11, t = 1.71, p = .089, but for older youth, increased prior narration was significantly related to increased learning, B = .31, t = 5.14, p < .001. As shown in Figure 3, the positive association between prior narration and reported learning was stronger among older participants.

Figure 3.

Simples slopes of the age by prior narration interaction: Reported learning for younger and older participants who engaged in less and more prior narration

Note. * p-value of slope < .05; n = 43 youth 1SD below mean age, n = 52 youth 1SD above mean age.

Finally, to probe the age by initial anger interaction, we examined perceptions of learning for younger (mean – 1 SD) and older (mean + 1 SD) participants who reported less (mean - 1 SD) and more (mean + 1 SD) anger when they initially recalled the angry memory. Tests of simple slopes revealed that whether initial anger was high or low mattered for reported learning for younger participants, B = .39, t = 4.82, p < .001, but not for older participants, B = .13, t = 1.54, p = .126 (see Figure 4). As shown, more anger-provoking experiences were associated with more reported learning for all participants, but as youth got older, reported learning for less anger-provoking experiences increased.

Figure 4.

Simples slopes of the age by intial anger interaction: Reported learning for younger and older participants who had lower and higher levels of initial anger

Note. * p-value of slope < .05; n = 43 youth 1SD below mean age, n = 52 youth 1SD above mean age.

Do reductions in anger relate to reported learning from the anger experience?

We examined whether reduced anger post-regulation and subsequent anger at re-exposure predicted youth’s reports of learning at the end of the initial session. We used OLS regression, and included age, gender, prior narration, and two dummy-coded condition variables so that we were looking at the effects of anger reduction over and above the relationships reported in earlier analyses. In addition, in keeping with prior analyses, we included interaction terms for age and the condition variables, and for age and prior narration ratings. Predictors were centered prior to computing interaction terms as per standard guidelines (Aiken & West, 1991).

The regression predicting learning from anger reduction and anger on immediate reexposure to the memory was significant, R² = .42, F(11,233) = 15.6, p <.001. Larger reductions in anger were significantly associated with higher ratings of perceived learning, β = .20, t = 3.46, p = .001, but so were higher ratings of anger on immediate reexposure, β = .23, t = 4.4, p < .001. As in prior analyses, older youth, and youth who had previously narrated the event reported greater learning, and interactions of age and the impact of narrative on learning, and age and the impact of prior narration on learning were also evident. An interaction of age with the extent of anger reduction was not significant, however, suggesting that the impact of anger reduction on self-reported learning did not vary across the age of our participants.

Narrate Participants Only: Exploring Links between Narrative Features and Learning

We assessed six narrative features: exploratory processing, factual elaboration, interpretive elaboration, positive meaning-making, resolution, and perspective-taking. An exploratory principle components analysis suggested that these features were primarily organized into two dimensions: elaboration (exploratory processing, factual elaboration, and interpretive elaboration) and positive resolution (positive meaning-making, resolution), with perspective-taking unrelated to either of the primary dimensions. We computed summary scores for these dimensions and left perspective-taking in our model as a single-item score. We then regressed learning on gender, age, prior narration, and the three narrative measures. Preliminary tests of interactions involving age and narrative features, and gender and narrative features, suggested there were no interactions involving age or gender in this sub-sample of the participants, so these interactions were not included in our regression models.

Overall, the included predictors accounted for a significant amount of variance in reported learning, R² = .42, F(6,114) = 12.98, p < .001. Youth whose narratives were more elaborated, β = .29, t = 3.4, p = .001, and whose narratives were less positively resolved, β = −.16, t = −2.2., p = .032 reported higher levels of learning. Of course, as expected from previous analyses of the full sample, older youth, β = .27, t = 3.3, p = .001, and youth who reported more previous narration, β = .35, t = 4.6, p < .001 reported more learning. No other effects were statistically significant.

Discussion

We set out to examine whether (1) narrating anger experiences promoted learning more than distraction or simply recalling the experience; (2) whether the impact of narration varied with the age of the child; (3) whether reductions in distress were important for promoting learning; and (4) whether learning was associated with the quality of the narratives participants created. Our findings suggest that narrating promoted learning more than mere remembering, but only among younger participants, supporting a modified version of our first hypothesis and contradicting our second. Consistent with our third hypothesis, larger reductions in distress from regulation were associated with more learning, but so was higher distress at reexposure, a finding we consider in more detail below. Finally, again broadly consistent with our fourth hypothesis, learning was higher for narratives with greater elaboration but less resolution. In addition, older youth and girls reported higher levels of learning in general, and age differences in learning were more marked for girls. Participants who reported narrating their anger event more extensively prior to our study also reported more learning. Next, we consider the implications of these findings for the broader literature on narrative, emotion regulation, learning, and development.

Effects of Narration, compared to distraction and a reexposure control, on Learning

To our knowledge this study is the first to examine narrative impacts on perceptions of learning separately from the features and contents of the narrative (e.g., McLean & Pratt, 2006), and to examine learning about the emotional experience, rather than about subsequent material (Davis & Levine, 2013). While it has often been assumed that narrating promotes learning (e.g., Laible & Thompson, 2002), explicit tests of this assumption are rare outside the adult literature on narrative identity, which focuses primarily on insights rather than learning per se (e.g., Weeks & Pasupathi, 2011) and the adult literature on post-traumatic growth, which also employs self-reports as a way to capture such growth (e.g., Tedeschi & Calhoun, 2004). Of course, post-traumatic growth is distinct in important ways from our child participants’ mundane and everyday anger experiences, and what they might learn from those events.

For younger participants, narrating improved learning over mere remembering, while for older participants, narrating was unrelated to learning. This pattern may emerge for a few reasons. First, prior narration was higher for older participants – and prior narration had positive impacts on learning. When youth have more extensively narrated the event prior to participating, it may be more difficult for our one-time laboratory-based narrative condition to impact learning. Second, older children may benefit from narration of anger to peers or in solitary conditions like a diary (see also Koeber et al., 2015; McLean & Breen, 2009; von Salisch & Saarni, 2011), whereas younger children may need the scaffolding of a basic, attentive adult listener to benefit from narration. This idea is consistent with our interaction effects involving age and experimental condition, but also helps make sense of the interaction of prior narration and age, in which only older youth reported higher learning with higher levels of prior narration. Further, older youth also reported higher learning for less intense angry experiences, suggesting general age-related increases in the capacity to learn from angry events that vary more widely in their impact – consistent with the age main effect on learning and with prior research in the area.

For older youth, distraction operated as we expected, in that engaging in distraction led to less reported learning about the anger event. More puzzling and unexpected were the findings about the impact of distraction on reported learning in younger children –where distraction promoted an increased sense of learning. For these children, the distraction and narrate tasks had in common engaging in a structured activity, and reductions in distress. These allow two different interpretations of the age and condition interaction. Speculatively, it may be that taking action in a structured and scaffolded way (as is provided by adult conversation partners and by games) promotes a sense of efficacy for younger children, and therefore a sense of having learned. By contrast, older youth may be able to productively re-engage with the anger experience via remembering it, as well as via narrating it, such that distraction undermines learning more so than the reexposure control. Alternatively, it may be that for younger children, the reduction in distress is what affords their sense that they have learned, whereas for older youth, the reduction in distress from distraction doesn’t provide a sense of having learned because it is divorced from the event itself. Evaluating these possibilities would require varied assessments of learning as well as of other potential mediating factors.

Our findings also suggest, consistent with earlier work by Davis and colleagues (Davis & Levine, 2013; Davis, 2016), that reducing anger effectively can promote learning. However, the fact that we also saw higher current anger related to higher learning suggests that residual emotion may be important. That is, when someone no longer feels any anger at all in contemplating a past experience, he or she may also perceive that they did not gain any insight from that event. Emotional distress marks experiences as warranting attention and memory (e.g., Christianson, 1997; Kensinger, Garoff-Eaton, & Schacter, 2007), so while reducing distress to a level that doesn’t interfere with cognition may promote learning, too large a reduction in distress may lead to indifference and reduce learning. Finally, we showed that, among participants assigned to the narrate condition, those whose narratives were more elaborative and more positively resolved reported more learning.

Given the benefits of narration – whether prior to our study or during our experiment – for perceived learning from anger events, it is worth underscoring the value of promoting youth narration of anger experiences, especially promoting high quality narration that is elaborated and resolved. Attentive and engaged listeners promote elaboration, and can (but do not always) promote positive resolution for people of all ages (e.g., Reese et al., 1993; McLean & Pasupathi, 2011; Nils & Rimé, 2012; Weeks & Pasupathi, 2011). For young children, it is important that listeners also scaffold psychological elaboration and resolution in ways that are developmentally appropriate (e.g., Recchia & Wainryb, 2014; Fivush et al., 2003; Bohanek et al., 2008). Existing work on anger expression in family contexts has suggested that among young children (8–12 year-olds) such expressions are mild and brief and primarily related to conflict between children and parents, and children and siblings (Sears, Repetti, Reynolds, & Sperling, 2014). This finding suggests that children are not especially likely to narrate anger, at least not in everyday family settings, consistent with our findings that prior narration and age were positively correlated. Our study’s findings strongly suggest that creating opportunities for youth to explore experiences of anger through narration can make space for their developing understandings of themselves and others.

Limitations and Future Directions

Like any study, ours has limitations. Among those were the fact that we have a single study (albeit with a large sample), and our laboratory session represents only one occasion on which participants might think about, seek to regulate, and/or learn from their anger experience. Future work may draw on the linguistic literature in looking at repeated narration over time (e.g,. Pennebaker & Seagal, 1999). We also worked with autobiographical experiences that are, of necessity, heterogeneous across our participants, although this limitation is simultaneously a strength, in that the experiences were meaningful and ecologically valid. Our assessment of learning, though novel within the narrative literature, was also limited to two single-item self-report ratings. This limitation reflects a dearth of options for looking at learning from heterogeneous everyday events, one worth addressing in future research. For example, participants could be asked to articulate what they have learned from experiences. Notably, the relationship between our assessment, and features of the narratives that are arguably also reflective of learning can be considered a kind of validity check. Additionally, we focused on the specific emotion of anger, but it is an open question whether these findings generalize to other emotions. In some cases, it makes sense to expect generalizability – for example, the idea that adult scaffolding might be important for younger children, the finding that both reducing distress and maintaining some distress are important for learning, and our findings about the importance of elaboration for learning. In other cases, such as the effects of prior narration, different emotions might operate differently. For example, adolescents might not always benefit from extensive prior narration of sadness, given the potential for rumination and the exacerbation of distress that may be especially likely during adolescence.

Despite the limitations, we believe the findings serve to underscore the importance of helping youth to narrate their experiences of anger. It is well established that this type of narration has benefits for connecting people with important others thereby creating social support (Bohanek et al., 2008; Pasupathi & Wainryb, 2018). Further, such narration can help to down-regulate the distress associated with anger (e.g., Marin et al., 2008; Pasupathi et al., 2017; Wainryb et al., 2018). This study, along with previous work examining parent-child co-narration (Bird & Reese, 2006; Fivush et al., 2003; Marin et al., 2008; Reese et al., 1993; Cleveland & Reese, 2005) extends the evidence for the benefits of narrating experiences beyond social and emotion regulation functions. Narration may be an important process by which we learn from our experiences of anger.