Online Emotion Regulation-Based Intervention for Reducing Anger: A Randomized Controlled Trial

Despite indications of the high prevalence of anger problems, such as reports from clinicians showing comparable numbers of clients with anger and anxiety issues (Lachmund et al., 2005), many individuals struggling with anger problems do not seek in-person treatment (Howie & Malouff, 2014). Various factors may contribute to this, including stigma, transportation challenges, or reduced availability of services. Online programs for depression have demonstrated comparable effectiveness to traditional face-to-face treatment (Selmi et al., 1990; Wright et al., 2005). Moreover, in general, the online format provides convenience, accessibility, and flexibility, enabling the delivery of psychological help to a broader population (Crisp et al., 2014; Griffiths et al., 2006; Lappalainen et al., 2014). Online programs that target anger, though few, show promise as an accessible and effective intervention for reducing problematic anger (DiBlasi & Sinn, 2024; Howie & Malouff, 2014; Osgood et al., 2020). In this study, we aimed to examine the short- and long-term effects of an online intervention in reducing both state- and trait-anger in adults with elevated levels of trait-anger. Individuals were instructed to engage in either reappraisal, acceptance, or neutral description of their activities (active-control condition). The authors posited the following hypothesis:

H1: The online reappraisal intervention will lead to a greater reduction in state- and trait-anger in the reappraisal-intervention group as compared to the control group;

H2: The online acceptance intervention will lead to a greater reduction in state- and trait-anger in the acceptance-intervention group as compared to the control group;

H3: The effects of the selected strategies on trait-anger (i.e., reappraisal, acceptance) will persist at 1-month follow-up.

Subsequent to institutional ethics approval, we recruited participants from the general community via social networks (Facebook, Instagram) from June 2022 to July 2023.

Adults (aged 18 years or older) who scored high on trait-anger (at or above the 75th percentile on the Trait-Anger Scale of the STAXI-2; Spielberger, 1988) were included in the study. Since we were interested in the overall impact of the intervention, participants were included in the analyses only if they completed the entire procedure. Exclusion criteria included current suicidal or homicidal ideation, and ongoing psychotherapy or psychiatric hospitalization, given that these severe mental health conditions could hinder the potential positive effects of an anger intervention. Thus, our final sample consisted of 78 Romanian participants (75 women, mean age = 36.22 years, SD = 9.38: Fig. 1). In terms of education, 16.7% had completed high school, 33.3% held a bachelor’s degree, 42.3% had a master’s degree, and 7.7% had obtained a doctoral degree.

Sample sizes were calculated in order to ensure sufficient power for detecting between-group differences in self-reports of angry affect at α = 0.05 and moderate to large effect sizes (i.e., 25–30; Cohen, 1992). Participants were randomly assigned to the active control and experimental groups (N = 30 in the acceptance group [mean age = 27.06 years, SD = 3.88; 28 female], N = 26 in the reappraisal group [mean age = 26.65 years, SD = 5.35; 23 female], and N = 22 in the control group [mean age = 26.63 years, SD = 3.88; 17 female]).

Once they completed all of the stages of the study, participants could choose their reward from course credit (if they were psychology students), gift cards, or access to an e-learning anger regulation module. No partial compensation was offered.

Informed consent was obtained from each participant.

The study was carried out online. Eligible participants completed the pre-test instruments, before being randomly assigned to one of three conditions: acceptance, reappraisal, or control. Participants in both experimental conditions (i.e., acceptance, reappraisal) received a two-hour online training on either acceptance or reappraisal. It was delivered by the first author of the study, a psychologist accredited by national institutions. Attendance at the online training was required to participate in the subsequent practice phase, described below. The Control group received only the practice phase. After completing the practice phase, participants received the post-test instruments, followed by the one-month follow-up. Trait-anger was measured at 3 time-points: prior to intervention (before participants were randomly assigned to one of the three conditions), after intervention (following the completion of the practice phase), and at 1-month follow-up. State-anger was measured 3 times during each of the 6 sessions, for a total of 18 assessments. In each session, the scale was administered at the beginning, after the anger-inducing exercise where participants wrote in detail about an angering event, and after the writing intervention where participants were instructed to use either reappraisal or acceptance or a neutral approach toward the angering event.

Linear mixed models were performed to investigate the effects of the interventions on state-anger using the statistical package JASP. Linear mixed models are appropriate for analyzing multilevel repeated assessment data because they account for the clustering of data by participants (level 2: Myers et al., 2012). Group (acceptance; reappraisal; control) and time (baseline; post-induction; post-instructions) were introduced as fixed effects and the random individual variation was accounted for. Models revealed a significant effect of the time (F(2, 75) = 46.20, p < 0.001), but no significant effect of the group, nor of the group and time interaction for state-anger. State-anger scores increased from baseline (time 1) to post-induction (time 2), indicating successful anger induction. Subsequently, a decrease was observed from post-induction (time 2) to post-instruction (time 3) across groups, suggesting reduced state-anger following instruction implementation (Fig. 2). State-anger reached its lowest level at Time 3, followed by Time 1, with the highest level observed at Time 2 (Supplementary Material).

A 3 (condition: acceptance, reappraisal, control) × 3 (time: pre-, post-, 1-month follow-up) mixed ANOVA was conducted to compare the effects of instructed ER and active-control condition on trait-anger. Table 1 contains means and standard deviation scores for all three-time measures in all groups (reappraisal, acceptance, and control).

There was a strong main effect for time, Wilks Lambda = 0.48, F(2, 74) = 38.97, p < 0.001, partial η2 = 0.51. However, the interaction between ER condition and time, and the main effect for the group were not significant. Post-hoc analysis (Bonferroni test) of the main effect for time showed that trait-anger significantly reduced from baseline (T1) to both post-intervention (T2) (mean difference: 3.99, SE = 0.47, p < 0.001), and 1-month follow-up (T3) assessments (mean difference: 4.39, SE = 0.57, p < 0.001). This pattern of decrease was observed across all the conditions being investigated. Thus, both the experimental groups and the active-control group experienced a significant reduction in trait-anger over time.

To check the efficacy of anger memory recall in increasing momentary anger, we calculated scores for state-anger after induction. The results demonstrated that the manipulation was successful.

Written responses from the writing task were analyzed to ensure adherence to the instructions for acceptance, reappraisal, and control.¹ Participants received feedback on the effectiveness of their strategy implementation, along with specific guidance for subsequent tasks.

Acceptance- and reappraisal-based online interventions contributed to significant changes in state- and trait-anger across time. However, the results did not support our hypothesis regarding the expectation that reappraisal and acceptance would lead to a greater reduction in both types of anger, as compared to simply acting neutrally about the event. One potential explanation is that the instructions provided to the active-control group may have served as guidelines for implementing distraction, as observed in another study (Denson et al., 2012). In this study participants in the control group, who were instructed to write freely about their thoughts, tended to engage in distraction. Similarly, participants in the active-control group may have focused their attention away from their emotions on neutral stimuli (Rusting & Nolen-Hoeksema, 1998), leading to a lower level of state-anger. Several studies support the effectiveness of distraction in reducing anger, largely due to its ease of implementation (Bushman et al., 2005; Denson et al., 2012; Millgram et al., 2019; Rusting & Nolen-Hoeksema, 1998). However, it has been discussed in the literature that distraction may only temporarily reduce anger without addressing its underlying source, suggesting its short-term effectiveness (Fabiansson & Denson, 2012). While distraction may indeed prevent individuals from processing the anger-inducing event, it also disrupts rumination (Larsen & Christenfeld, 2011; Neumann et al., 2004; Pedersen et al., 2011) which has consistently been shown to increase anger (Offredi et al., 2016; Ray et al., 2008; Rusting & Nolen-Hoeksema, 1998). Thus, distraction may have a sustained effect on anger by preventing rumination. Additionally, given that individuals tend to resort to distraction when prompted to use neutral or spontaneous ER strategies (Denson et al., 2012), this suggests that distraction is a common anger regulation strategy across individuals, and warrants further research.

Another potential explanation for the null outcomes regarding both state- and trait-anger may lie in the limited duration of the intervention, which may not have enabled a broad shift in perspectives (reappraisal) or a comprehensive acceptance and tolerance of negative feelings. While we wanted to investigate an online intensive time-limited intervention, deeper internalization of reappraisal and acceptance likely requires a more extended intervention time. Although the initial two-hour training provided participants with explanations and examples of using reappraisal and acceptance, and applying these strategies to angering situations that participants shared, some key elements may have been missing. For instance, the reappraisal training focused solely on pragmatic disputation, but not logical and empirical, largely due to time constraints. Furthermore, while participants were given direct instructions during the practice sessions, incorporating more Socratic questioning might have encouraged a deeper cognitive shift specific to reappraisal and a more pervasive adoption of acceptance. Similarly, acceptance might require more time to fully integrate into one’s ER strategies repertoire and have a slower down-regulating effect on anger (see Germain & Kangas, 2015). Moreover, particularly if targeting trait-anger, which is generally considered to be a more stable tendency to experience anger (Veenstra et al., 2018), longer intervention time may be more appropriate.

Despite the potential explanations, the null outcomes suggest that the observed changes in both trait- and state-anger that across time in all groups may be attributed to the participants’ expectations or to the general sense of support from the research team, rather than the specific strategy they were instructed to use. In future studies, researchers could investigate the participants’ expectations and perceptions of support to determine the extent to which these factors may contribute to the observed reduced anger. Given that common factors such as empathy and the therapeutic relationship often exhibit stronger correlations with psychotherapeutic success than specific therapy techniques (Lambert & Barley, 2001; Lambert & Ogles, 2004), understanding their contribution is essential for refining anger management approaches. Interventions may benefit from incorporating elements that enhance perceived support and motivation, rather than relying solely on teaching new strategies. Additionally, understanding the role of expectations could improve how therapeutic techniques are introduced and applied in practice. A key question that arises, therefore, is whether we should continue investing in research on the effectiveness of specific strategies and skill development. Evidence suggests that specific therapy techniques do contribute to psychological outcomes, accounting for approximately 15% of the variance (Lambert & Barley, 2001). Moreover, recent research highlights the crucial role of emotion regulation flexibility–shifting regulatory efforts based on contextual demands–as a central factor in mental health (Aldao et al., 2015; Bonanno & Burton, 2013). Maintaining a broad repertoire of strategies and having the ability to use a diverse range of strategies are indices of this construct and are associated with lower long-term distress (Bonanno et al., 2004; English & Eldesouky, 2020). Thus, learning to use various ER strategies may facilitate ER flexibility, enabling individuals to regulate their emotions by adapting to contextual features (Kalokerinos & Koval, in press).

Several methodological limitations warrant consideration when interpreting the current findings. The reliance on self-reported assessments represents a limitation. However, due to the online delivery of the intervention, utilizing self-reported measures was a more feasible option. Future studies should consider integrating a multimodal assessment approach to anger, including clinical interviews or observer-reported assessments, whether in an online or in-person format.

For increased evaluation reliability, validated measures with multiple items were used. However, given the multiple times assessments of anger, participants may have experienced fatigue and distraction when filling out the questionnaire, potentially impacting response accuracy (Table 2).

Another important issue to address is the absence of a direct manipulation check of the ER strategies, which limits the interpretation of our findings. In future studies, investigators could consider using multiple-item measures to assess specific ER strategies employment.

The gender imbalance (75 of the 78 participants were women) is another limitation of the current study. One potential explanation is related to gender roles: women are generally less expected to express anger (Fischer & Evers, 2011), and therefore, they may have been more motivated to participate in the two-week intervention. In future studies it would be wise for investigators to place extra effort on recruiting more gender-balanced samples to ensure that the findings are generalizable to both men and women. Additionally, exploring the influence of gender roles on treatment-seeking behavior and engagement may provide valuable insights for tailoring interventions to better meet the personal needs of different individuals.

While a 1-month follow-up was a strength of the study, extending the follow-up period to include longer-term assessments would provide additional insights. This could involve 3-, 6-month follow-up assessments or evaluating daily use of ER strategies and associated emotional outcomes (Fig. 3).

Lastly, the high drop-out rate (32.17%) resulted in a smaller sample size, limiting our ability to detect small to medium effect sizes and only allowing for the identification of large effects for potential between-group differences, which might be overestimated by Cohen’s guidelines (Lovakov & Agadullina, 2021). Thus, considering that effects are often modest in psychological intervention research, particularly when comparing active interventions to control conditions, the final sample size represents a significant constraint in detecting subtle but clinically significant differences.

Participants may have felt bored by the experimental task itself, which involved applying the same instructions, delivered in a text format. Future ER strategies-based interventions should consider more engaging content, such as interactive exercises, with automatic feedback, explanatory videos and images.