Authors: Gabriela Viorela Pop, Diana-Mirela Nechita, Andrei C. Miu, Aurora Szentágotai-Tătar
Categories: Article, Anger, Emotion regulation, Psychopathology, Meta-analysis, Quality of life, Human behaviour
Source: Scientific Reports
Authors: Gabriela Viorela Pop, Diana-Mirela Nechita, Andrei C. Miu, Aurora Szentágotai-Tătar
The hypothesis that maladaptive emotion regulation plays a central role in anger problems has driven the development of cognitive-behavioral interventions for anger and has stimulated a wealth of studies. However, this work is heterogeneous, drawing on multiple theories from clinical psychology and affective science, and focusing on multiple dimensions of emotion regulation. In addition, previous findings have not always been consistent. The present meta-analysis aimed to characterize associations between anger and emotion regulation strategies (i.e., avoidance, acceptance, distraction, reappraisal, rumination, suppression). A systematic search was conducted, and 81 studies (115 effect sizes) were included in the analysis. The results indicated consistent positive associations between anger and avoidance, rumination, and suppression, and consistent negative associations between anger and acceptance, and reappraisal. The relation between anger and distraction could not be analyzed. Heterogeneity was large in all analyses, and multiple differences between studies, including type of anger, clinical status, criminal status, type of culture, study quality, and sample sex distribution, influenced the effect sizes. Overall, the present findings support the view that anger is consistently associated with the differential use of multiple emotion regulation strategies and suggest methodological improvements for future studies.
Anger is a common emotional experience in everyday life^1,2^. It unfolds when individuals believe that they will not attain a goal due to the actions of another person, and that the situation can be changed^3^. The putative function of anger is to promote confrontation with others, which can be adaptive in certain contexts^4^, but it can also escalate to aggressive behavior^5,6^. When anger is too frequent or too intense, it is associated with a range of antisocial behaviors including domestic violence^7^, childhood maltreatment^8^, aggressive driving^9^ and even homicide^10^. Furthermore, increased anger has been documented in multiple mental disorders, being associated with greater symptom severity, greater comorbidity and lower quality of life^11–13^.
According to an influential perspective, maladaptive emotion regulation is a central mechanism underlying problematic anger and its negative implications^14,15^. This view has driven the development of cognitive-behavioral interventions in anger^16–18^ and has stimulated a wealth of studies^15^. However, this work is extremely heterogeneous, drawing on multiple theories from clinical psychology and affective science, and focusing on multiple dimensions of emotion regulation.
One of the theoretical frameworks that have focused on the role of emotion regulation in anger and aggression has been articulated by Roberton and coworkers^15^. They argue that aggression is driven by maladaptive regulation of negative affect and particularly anger. According to this view, there is a continuum of maladaptive emotion regulation “styles” underlying aggression, ranging from emotion under-regulation to emotion over-regulation. Emotion under-regulation refers to failures to regulate negative affect (e.g., anger) in a sufficient manner so as to maintain goal-directed behavior. Over-regulation relies on attempts to block negative affect altogether, either by avoiding unpleasant emotional experiences or the situations that may trigger it, or by suppressing its expression. These maladaptive emotion regulation styles involve a variety of strategies, which are not inherently maladaptive, but become so when used improperly (e.g., irrespective of situational demands), making them inefficient, or when employed to avoid unpleasant emotional experiences^15^.
Anger avoidance is also in the focus of Gardner and Moore’s theoretical framework^14^. Rumination, an emotion regulation strategy that involves repetitive thinking about negative emotions^19^, is viewed here as a cognitive form of anger avoidance, whereas aggressive behavior is conceptualized as a behavioral form of anger avoidance. When used to escape anger, both strategies are hypothesized to be maintained through negative reinforcement^14^.
Emotion regulation is an important part of the integrative cognitive model of trait anger^20^. This theory argues that certain individuals are prone to automatic hostile interpretations and consequently, more frequent anger. Rumination, which is conceptualized in this model as a form of attentional dysregulation, can further contribute to the amplification of anger. In contrast, controlled processes can buffer anger reactions, through the reappraisal of the situation as non-hostile (for a similar role in aggressive behavior, see ^21^) and through the effortful interruption of rumination and distraction from hostile thoughts. Effortful control can also involve suppressing anger and aggressive behavior.^20^.
Overall, these theoretical models assume that maladaptive emotion regulation is central to problematic anger. They also attempt to identify the ways in which emotion regulation can be maladaptive (e.g., under- or over-regulation), and illustrate the variety of strategies involved in both the upregulation (e.g., rumination, suppression) and the downregulation (e.g., reappraisal, distraction) of anger.
The focus on multiple emotion regulation strategies in current research dates back to pioneering clinical interventions and basic studies on anger. For example, early work on cognitive-behavioral interventions^22^ has highlighted the special role of reappraisal in anger regulation. Reappraisal is an emotion regulation strategy that involves reinterpreting the situation in order to modulate its emotional impact^23^, and is a central skill targeted by Novaco’s stress inoculation training in anger^22^ and other cognitive-behavioral interventions^17,18^. The efficiency of these interventions has been consistently supported^17,18,24^, well before the impact of using reappraisal on reducing anger was corroborated by experimental^25^ and individual differences^26^ (e.g., dispositional traits) research. In a similar manner, experimental studies^27^ have investigated the potential effect of acceptance (i.e., leaving emotion unfold naturally) in reducing anger, in light of the central role attributed to this emotion regulation strategy in acceptance- and mindfulness-based interventions^28,29^.
Another emotion regulation strategy linked to anger in early work is expressive suppression. This strategy involves inhibiting the expression of emotion, and was found in early experimental work to increase cardiovascular activity (e.g., heart rate, systolic blood pressure) in anger^30^. Coined as “anger-in”, anger suppression has also been assessed as an individual difference^31–33^ and has been associated with a host of emotional and physiological characteristics^34^.
In studies examining anger, rumination was associated with both increased aggressive behavior and slower physiological recovery (i.e., prolonged cardiovascular reactivity) following anger-inducing events^35,36^. Notably, rumination was compared either with distraction^35^, following in the footsteps of early studies on the response styles theory in depression^37^, or reappraisal^25^. Similarly, other studies compared between reappraisal, acceptance and suppression, and found that former strategy was more efficient in reducing anger compared to the latter two strategies^27^.
Roberton et al. (p. 79)^15^ have argued that the concept of emotion regulation is clearly cornerstone in current research on problematic anger, but that “unless a cohesive stream of research is established, it may become difficult to draw conclusive findings from different studies.” The present meta-analysis set out to offer an integrative perspective on current evidence, by identifying consistent associations between anger and multiple emotion regulation strategies. We focused on six emotion regulation strategies and, in light of current perspectives on the relative efficiency of emotion regulation strategies^38^, we expected that some (e.g., acceptance, distraction, reappraisal) showed a negative and others (e.g., avoidance, rumination, suppression) showed a positive association with anger.
To reduce heterogeneity related to methodological differences between studies in adults vs. other age categories (e.g., children), we restricted the present meta-analysis to adults. Also, we focused on subjective measures of anger. Another element of our approach, designed to limit heterogeneity, was our focus on the habitual and spontaneous use of emotion regulation strategies. This focus was reflected in self-report questionnaires that assessed both the global use of these strategies and their spontaneous use in specific contexts, such as through ecological momentary assessment. Because these measures did not overlap with the instructed use of emotion regulation^39^, we decided to exclude experimental studies in which emotion regulation strategies were manipulated.
Another aim of the present meta-analysis was to uncover differences between studies that could explain heterogeneity in previous results on anger and emotion regulation. First, we compared effect sizes in subgroups of studies that focused on trait vs. state anger. Considering that trait anger is a relatively stable individual difference^33^, whereas state anger varies across contexts (which may or may not be specified), we expected the former to be more consistently associated with emotion regulation.
Clinical status was also examined as potential moderator considering that problematic anger has been associated with multiple mental disorders^40^. Here, we also considered the distinction between studies that focused on clinical samples (i.e., patients diagnosed with a mental disorder) and those that selected individuals for high levels of anger, involving clinically relevant behavioral dysfunctions^41^. We expected higher effect sizes of the associations between anger and emotion regulation in studies on clinical samples and samples with clinically relevant levels of anger, compared to general samples.
Another potential moderator that we considered was criminal status, in light of studies showing that both problematic anger^42^, and maladaptive emotion regulation^43^ are increased in offenders compared to non-offenders.
Cultural differences were also examined as a potential moderator. Individualist cultures emphasize the role of emotions in individual wellbeing, whereas collectivist cultures prioritize social convention^44^, and these differences influence the use of emotion regulation and its consequences. For instance, in individualist^45^, but not in collectivist cultures^46^, suppression is associated with increased anger. In cultures where anger expression is incongruent with social norms, rumination might also be increased^47^. Therefore, we contrasted subgroups of studies in individualist vs. collectivist cultures.
Sample differences were also investigated as potential moderators, particularly sex distribution and age category. Previous studies on sex differences in the use of emotion regulation strategies such as suppression^48,49^ and rumination^50,51^, are inconsistent, but this issue has remained open. Previous work has also investigated potential differences in anger regulation between younger and older adults, also with apparently inconsistent results on strategies such as suppression^52,53^ and reappraisal^54,55^.
Finally, we considered the possibility that methodological differences explained heterogeneity in previous studies. Study design (i.e., cross-sectional vs. longitudinal) and study quality were examined, with the expectation that longitudinal and higher-quality studies showed more consistent associations between anger and emotion regulation.
As illustrated in Fig. 1, the search returned 9262 articles. After removing 4256 duplicates, a total of 5006 articles were screened based on title and abstract, and 4808 articles were excluded for not meeting the inclusion criteria. The remaining 198 articles were evaluated full-text, and 114 were excluded based on the inclusion and exclusion criteria.Fig. 1PRISMA flowchart showing the selection of studies.
A total of 73 studies met the eligibility criteria, and an additional 11 were also eligible, but did not provide enough data for effect size estimation. The authors of studies in the latter category were contacted and they provided data from 8 studies. Information on missing group characteristics was also requested for 14 of the 82 studies, and 3 authors responded.
Table 1 shows the characteristics of studies. Eighty-one studies were included in the analysis and provided 115 effect sizes on the association between anger and one of the emotion regulation suppression (k = 41), rumination (k = 37), reappraisal (k = 12), acceptance (k = 12), and avoidance (k = 12). A single study reported data on the association between anger and distraction and was discarded from analyses. Most of these studies (93.82%) were cross-sectional, and the rest (k = 5) were longitudinal. Thirteen studies reported data from clinical samples, and the rest involved healthy participants.Table 1Characteristics of studies included in the meta-analysis.Study referenceNMean age% womenCountryIndividualist/CollectivistClinical statusCriminal statusType of angerAnger scaleType of ER strategyER strategy scaleEffect size (r)Akinci and Atintas^56^49632.872.4Turkeycollectivistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGAvoidanceCoSS/Avoidance0.11Alcázar, et al.^57^9521.6769.56Mexicocollectivistclinical angernon-offendersTraitML-STAXI/Trait ANGSuppressionML-STAXI/ANG IN0.50Alcázar-Olán, et al.^58^37620.4860.9Mexicocollectivistclinical angernon-offendersTraitML-STAXI/Trait ANGSuppressionML-STAXI/ANG IN0.53Anestis, et al.^59^20018.3268.5USAindividualistnon-clinicnon-offendersTraitBPAQ/ANGRuminationARS0.40Arslan^60^46819.7555.12Turkeycollectivistnon-clinicnon-offendersTraitSTAXI/Trait ANGAvoidanceCWSS/Avoidance-0.01Balsamo^61^35332.665.2Italyindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGRuminationPadua Inventory/Tendency to Doubt and to Ruminate0.48Berry, et al.^62^23319.266.5USAindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGRuminationDRS0.60Besharat^63^93322.558.41Irancollectivistnon-clinicnon-offendersTraitTehran MAS/Trait ANGRuminationARS0.59Besharat^63^93322.558.41Irancollectivistnon-clinicnon-offendersStateTehran MAS/State ANGRuminationARS0.55Boersma-van Dam, et al.^64^11143.880.2Belgium + the Netherlandsindividualistnon-clinicnon-offendersEvent-related State1-item AngerRuminationCERQ/Rumination0.22Borders, et al.^161^ (Study 1)17019.766USAindividualistnon-clinicnon-offendersTraitBPAQ/AngerRuminationRRQ0.40Borders, et al.^161^ (Study 2)13431.860mixedindividualistnon-clinicnon-offendersTraitBPAQ/AngerRuminationRRQ0.47Borders and Liang^65^ (Sample 1)200--USAindividualistnon-clinicnon-offendersTraitBPAQ/AngerRuminationARS0.38Borders and Liang^65^ (Sample 2)150--USAindividualistnon-clinicnon-offendersTraitBPAQ/AngerRuminationARS0.47Borders and Lu^66^12819.4878USAindividualistnon-clinicnon-offendersTraitBPAQ/AngerRuminationARS0.45Borders and Wiley^67^ (Sample 1)15522.4946.7USAindividualistnon-clinicnon-offendersEvent-related StateAnger 7-item 7-point LikertRuminationRIOS-Modified0.57Borders and Wiley^67^ (Sample 2)46419.5678USAindividualistnon-clinicnon-offendersEvent-related StateAnger 7-item 7-point LikertRuminationRIOS-Modified0.28Borders and Wiley^67^ (Sample 3)21119.45100USAindividualistnon-clinicnon-offendersEvent-related StateAnger 7-item 7-point LikertRuminationRIOS-Modified0.43Brassard, et al.^68^302350Canadaindividualistclinical angernon-offendersStateIAEC/State ANGSuppressionSTAXI/ANG IN0.34Breen and Kashdan^69^17021.2975.3USAindividualistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGAvoidanceAAQ-II0.37Breen and Kashdan^69^17021.2975.3USAindividualistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGSuppressionSTAXI-2/ANG IN0.32Burns, et al.^70^10546.348.6USAindividualistnon-clinicnon-offendersStateAnger 3-item 9-point LikertSuppressionSuppression 4-item 9-point Likert0.38Carr^71^ (Time 1)16469.1573.17USAindividualistnon-clinicnon-offendersTraitAnger 3-itemReappraisalStrategies for coping with widow(er)hood/Positive reframing-0.05Carr^71^ (Time 2)16469.1573.17USAindividualistnon-clinicnon-offendersTraitAnger 3-itemReappraisalStrategies for coping with widow(er)hood/Positive reframing0.03Cheung and Park^72^36519.355.9USAindividualistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGSuppressionSTAXI-2/ANG IN0.35Cho, et al.^73^60541.3355.2USAindividualistnon-clinicnon-offendersEvent-related StateAnger 3-item 7-point LikertAcceptanceAcceptance 3-item 7-point Likert0.04Cho, et al.^73^60541.3355.2USAindividualistnon-clinicnon-offendersEvent-related StateAnger 3-item 7-point LikertAvoidanceAvoidance 3-item 7-point Likert-0.10Cludius, et al.^74^4832.4650Germanyindividualistclinicnon-offendersTraitSTAXI-2/Trait ANGAcceptanceDERS/Nonacceptance-0.59Cludius, et al.^74^4832.4650Germanyindividualistclinicnon-offendersTraitSTAXI-2/Trait ANGSuppressionSTAXI-2/ANG IN0.46Coleman, et al.^75^ (Time 1)9632.40USAindividualistnon-clinicnon-offendersEvent-related StateAnger 3-item 6-point LikertRumination9-item DRS0.47Coleman, et al. ^75^ (Time 2)8032.40USAindividualistnon-clinicnon-offendersEvent-related StateAnger 3-item 6-point LikertRumination9-item DRS0.61Conrad, et al.^76^ (Sample 1)9435.7562.8Germanyindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGSuppressionSTAXI/ANG IN0.24Conrad, et al.^76^ (Sample 2)32138.8358.9Germanyindividualistclinicnon-offendersTraitSTAXI/Trait ANGSuppressionSTAXI/ANG IN0.39Consedine, et al.^77^136159.3100USAindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGSuppressionPPQ0.24Crisan and Nechita^78^20325.5584.7Romaniacollectivistclinicnon-offendersTraitSTAXI-2/Trait ANGAvoidanceAAQ-II0.16Crisan and Nechita^78^20325.5584.7Romaniacollectivistclinicnon-offendersTraitSTAXI-2/Trait ANGRuminationARS0.45da Costa, et al.^79^9831.328.6Portugalcollectivistnon-clinicnon-offendersStateSTAXI/State ANGReappraisalERQ/R0.07da Costa, et al.^79^9831.328.6Portugalcollectivistnon-clinicnon-offendersTraitSTAXI/Trait ANG-TempReappraisalERQ/R0.01da Costa, et al.^79^9831.328.6Portugalcollectivistnon-clinicnon-offendersStateSTAXI/State ANGSuppressionERQ/S0.04da Costa, et al.^79^9831.328.6Portugalcollectivistnon-clinicnon-offendersTraitSTAXI/Trait ANG-TempSuppressionERQ/S-0.03Dahlen and Deffenbacher^80^8619.376.74USAindividualistclinical angernon-offendersEvent-related StateAnger 7-day logSuppressionSTAXI/ANG IN0.08Dahlen and Deffenbacher^80^8619.376.74USAindividualistclinical angernon-offendersTraitSTAXI/Trait ANGSuppressionSTAXI/ANG IN0.04Deffenbacher, et al.^81^48-45.83USAindividualistclinical angernon-offendersEvent-related StateSTAXI/State ANGSuppressionSTAXI/ANG IN0.17Deffenbacher, et al.^81^48-45.83USAindividualistclinical angernon-offendersTraitSTAXI/Trait ANGSuppressionSTAXI/ ANG IN0.42Dennison and Stewart^82^22225.576Australiaindividualistnon-clinicnon-offendersTraitList of descriptors for Anger/1-factor AngerRuminationList of descriptors/1-factor Rumination0.14Donahue, et al.^83^2354.830USAindividualistclinical angernon-offendersTraitAnger DARAvoidanceAAQ-II0.79Emmerich, et al.^84^39745.8390.5German countriesindividualistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGSuppressionSTAXI-2/ANG IN0.20Franco, et al.^85^3641.27100Spainindividualistnon-clinicnon-offendersTraitPOMS/AngerAvoidanceAAQ0.14Fresnics and Borders^86^20119.4878USAindividualistnon-clinicnon-offendersStateSTAXI-2/State ANGRuminationARS0.33Garofalo, et al.^87^ (Sample 1)19738.880Italyindividualistnon-clinicnon-offendersTraitBPAQ/AngerAcceptanceDERS/Nonacceptance-0.33Garofalo, et al.^87^ (Sample 2)15341.780Italyindividualistnon-clinicoffendersTraitBPAQ/AngerAcceptanceDERS/Nonacceptance-0.36Garofalo, et al. ^88^ (Sample 1)24538.90Italyindividualistnon-clinicnon-offendersTraitBPAQ/AngerAcceptanceDERS/Nonacceptance-0.39Garofalo, et al.^88^ (Sample 2)22140.90Italyindividualistnon-clinicoffendersTraitBPAQ/AngerAcceptanceDERS/Nonacceptance-0.44Garofalo, et al.^89^ (Sample 1)32437.870Italyindividualistnon-clinicnon-offendersTraitBPAQ/AngerAcceptanceDERS/Nonacceptance-0.38Garofalo, et al.^89^ (Sample 2)39740.260Italyindividualistnon-clinicoffendersTraitBPAQ/AngerAcceptanceDERS/Nonacceptance-0.38Halperin and Gross^90^201-49.75Israelindividualistnon-clinicnon-offendersEvent-related StateAnger 1-item 6-point LikertReappraisal3-item abbreviated version of ERQ/R0.08Han, et al.^91^371-80.32Chinacollectivistnon-clinicnon-offendersTraitNAS/part BRuminationRRS0.29Haukkala^92^ (Sample 1)1402460Finlandindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGSuppressionSTAXI/ANG IN0.30Haukkala^92^ (Sample 2)164245.5100Finlandindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGSuppressionSTAXI/ANG IN0.22Jordaan and Hesselink^93^243250South Africaindividualistnon-clinicoffendersTraitBPAQ/AngerAvoidanceCSI/Avoidance0.26Kale and Gedik^94^15627.2510.3Turkeycollectivistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGReappraisalERQ-R-0.14Kale and Gedik^94^15627.2510.3Turkeycollectivistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGSuppressionSTAXI-2/ANG IN0.51Kitayama, et al.^95^ (Sample 1)105058.0455.04USAindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGSuppressionSTAXI/ANG IN0.49Kitayama, et al.^95^ (Sample 2)36255.4759.11Japancollectivistnon-clinicnon-offendersTraitSTAXI/Trait ANGSuppressionSTAXI/ANG IN0.43Kong, et al.^96^57322.2649.56Chinacollectivistnon-clinicnon-offendersTraitSTAXI/ANGRuminationRumination 7-item 5-point Likert0.36Kosson, et al.^97^ (Sample 1)12240.1753.5Italyindividualistnon-clinicnon-offendersTraitDES-IV Trait AngerSuppressionSTAXI-2/ANG IN0.32Kosson, et al.^97^ (Sample 2)16141.490Italyindividualistnon-clinicoffendersTraitDES-IV Trait AngerSuppressionSTAXI-2/ANG IN0.53Lee, et al.^98^18943.0156.6Germanyindividualistnon-clinicnon-offendersTraitAEQ-Teacher/ANGReappraisalERQ/R0.04Lee, et al.^98^18943.0156.6Germanyindividualistnon-clinicnon-offendersTraitAEQ-Teacher/ANGSuppressionERQ/S0.05Li and Xia^99^ (Time 1)120918.9262Chinacollectivistnon-clinicnon-offendersTraitSTAXI/Trait ANGRuminationDAQ/Anger Rumination0.50Li and Xia^99^ (Time 2)120918.9262Chinacollectivistnon-clinicnon-offendersTraitSTAXI/Trait ANGRuminationDAQ/Anger Rumination0.49Li and Xia^99^ (Time 3)120918.9262Chinacollectivistnon-clinicnon-offendersTraitSTAXI/Trait ANGRuminationDAQ/Anger Rumination0.44Li, et al.^100^958-62.7Chinacollectivistnon-clinicnon-offendersEvent-related StateAnger 3-item 5-point LikertReappraisalReappraisal 4-item 5-point Likert-0.12Li, et al.^100^958-62.7Chinacollectivistnon-clinicnon-offendersEvent-related StateAnger 3-item 5-point LikertSuppressionSuppression 2-item 5-point Likert0.06Martin and Dahlen^26^36220.4679USAindividualistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGReappraisalCERQ/Positive Reappraisal-0.22Martin and Dahlen^26^36220.4679USAindividualistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGRuminationCERQ/Rumination0.26Martin and Dahlen^26^36220.4679USAindividualistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGSuppressionSTAXI/ANG IN0.29Massa, et al.^101^27133.1436.6USAindividualistnon-clinicoffendersTraitBPAQ/AngerRuminationRRQ0.50Mathes, et al.^102^11942.2648.7USAindividualistclinicnon-offendersTraitBPAQ/AngerRuminationRRS0.45McGonigal and Dixon-Gordon^103^3029.870USAindividualistnon-clinicoffendersTraitSTAXI-2/Trait ANGAcceptanceDERS/Nonacceptance-0.46Monaci and Veronesi^104^ (Sample 1)88-100Italyindividualistnon-clinicnon-offendersEvent-related StateSTAXI-2/State ANGSuppressionSTAXI/ANG IN0.34Monaci and Veronesi^104^ (Sample 2)92-0Italyindividualistnon-clinicnon-offendersEvent-related StateSTAXI-2/State ANGSuppressionSTAXI/ANG IN0.19Moron and Biolik-Moron^105^ (Study 1)21829.645.87Polandindividualistnon-clinicnon-offendersTraitBPAQ/AngerSuppressionCECS/Anger suppression-0.17Moron and Biolik-Moron^105^ (Study 2)10327.569.9Polandindividualistnon-clinicnon-offendersEvent-related StateSTAXI-2/State ANGAvoidanceAAQ-II0.08Moron and Biolik-Moron^105^ (Study 2)10327.569.9Polandindividualistnon-clinicnon-offendersEvent-related StateSTAXI-2/State ANGRuminationDAQ/Anger Rumination-4 items selected0.39Moron and Biolik-Moron^105^ (Study 2)10327.569.9Polandindividualistnon-clinicnon-offendersEvent-related StateSTAXI-2/State ANGSuppressionSTAXI-2/ANG IN0.18Navarro-Haro, et al.^106^10927.38100Spainindividualistclinicnon-offendersStateSTAXI-2/State ANGReappraisalERQ/R-0.15Navarro-Haro, et al.^106^10927.38100Spainindividualistclinicnon-offendersTraitSTAXI-2/Trait ANGReappraisalERQ/R-0.41Navarro-Haro, et al.^106^10927.38100Spainindividualistclinicnon-offendersStateSTAXI-2/State ANGSuppressionERQ/S-0.12Navarro-Haro, et al.^106^10927.38100Spainindividualistclinicnon-offendersTraitSTAXI-2/Trait ANGSuppressionERQ/S-0.14Offredi, et al.^107^4522.7966.66Italyindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGRuminationARS0.28Park, et al.^108^23353.5967.38USAindividualistnon-clinicnon-offendersTraitAnger 1-item 5-point LikertSuppressionSTAXI/ANG IN-0.14Peled and Moretti^51^22619.768.58USAindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANG-TempRuminationSARI0.33Peters, et al. ^109^82319.2570.6USAindividualistnon-clinicnon-offendersTraitBPAQ/AngerRuminationARS0.44Prikhidko and Swank^110^227-100USAindividualistnon-clinicnon-offendersTraitPAS/Anger experienceReappraisalRevised ERQ-9-0.27Prikhidko and Swank^110^227-100USAindividualistnon-clinicnon-offendersTraitPAS/Anger experienceSuppressionRevised ERQ-90.00Quan, et al.^111^60020.5151.67Chinacollectivistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGRuminationARS0.49^Salguero et al.112^ (Study 1)31435.4980.6Australiaindividualistnon-clinicnon-offendersTraitDAR-RRuminationARS0.61^Salguero et al.112^ (Study 2)63336.6256.9Spainindividualistnon-clinicnon-offendersTraitDAR-RRuminationARS0.57Sander, et al.^113^16543.794USAindividualistnon-clinicnon-offendersTraitPROMIS AngerSuppressionTBI-CareQQL Emotional Suppression0.29Schröder‐Abé, et al.^114^7923.7369.62Germanyindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGSuppressionSTAXI/ANG IN0.08Shallcross, et al.^115^ (Time 1)34041.3248USAindividualistnon-clinicnon-offendersTraitAnger 2-item 5-point LikertAcceptanceKIMS/Acceptance-0.29Shallcross, et al.^115^ (Time 2)28041.3248USAindividualistnon-clinicnon-offendersEvent-related StateAnger 2-item 5-point LikertAcceptanceKIMS/Acceptance-0.18Shallcross, et al. ^115^ (Time 3)24441.3248USAindividualistnon-clinicnon-offendersStateAnger 2-item 5-point LikertAcceptanceKIMS/Acceptance-0.25Shallcross, et al.^115^ (Time 4)23941.3248USAindividualistnon-clinicnon-offendersTraitAnger 2-item 5-point LikertAcceptanceKIMS/Acceptance-0.15Smeijers, et al.^116^41320.0482.6the Netherlandsindividualistnon-clinicnon-offendersStateSTAXI/State ANGRuminationDAQ/Anger rumination0.31Smeijers, et al.^116^41320.0482.6the Netherlandsindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGRuminationDAQ/Anger rumination0.47Sofia and Cruz^117^26921.7328Portugalcollectivistnon-clinicnon-offendersTraitCAAS/AngerReappraisalERSCS/ Reappraisal-0.13Sofia and Cruz^117^26921.7328Portugalcollectivistnon-clinicnon-offendersTraitCAAS/AngerRuminationDAQ/Anger rumination0.31Sofia and Cruz^117^26921.7328Portugalcollectivistnon-clinicnon-offendersTraitCAAS/AngerSuppression‘White Bear’' Suppression Inventory0.20Solak, et al.^118^13121.667.17Turkeycollectivistnon-clinicnon-offendersTraitAnger 3-item 7-point LikertSuppressionERQ/S-0.08Stover, et al.^119^37833.590USAindividualistnon-clinicoffendersTraitMAI/AngerAcceptanceDERS/Nonacceptance-0.35Sullivan and Kahn^120^9720.5953.6USAindividualistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGSuppressionERQ/S0.13Sun, et al.^121^3025-63.3Chinacollectivistnon-clinicnon-offendersStateAnger 1-item 5-optionReappraisalERQ/R-0.14Sun, et al.^121^3025-63.3Chinacollectivistnon-clinicnon-offendersStateAnger 1-item 5-optionSuppressionERQ/S-0.06Thomsen, et al.^122^12623.3260.31Denmarkindividualistnon-clinicnon-offendersTraitPOMS-37/ANGRuminationECQ-R0.23Toohey, et al.^123^7952.17.59USAindividualistclinicnon-offendersStateSTAXI-2/State ANGAvoidanceAAQ-II0.22Toohey, et al.^123^7952.17.59USAindividualistclinicnon-offendersTraitSTAXI-2/Trait ANGAvoidanceAAQ-II0.12Torres-Marín, et al.^124^16428.853.04Spainindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGSuppressionSTAXI/ANG IN0.26Trew and Alden^125^36319.5573.3Canadaindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGRuminationRRS-Brooding0.33Trew and Alden^125^36319.5573.3Canadaindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGRuminationRRS-Reflective Pondering0.14Trew and Alden^125^36319.5573.3Canadaindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGSuppressionSTAXI/ANG IN0.44Tull, et al.^126^11326.410USAindividualistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGAvoidanceAAQ0.19van Middendorp, et al.^127^33147100the Netherlandsindividualistnon-clinicnon-offendersEvent-related StateAnger 2 item 5-point LikertSuppressionSECS/ANG IN0.14Velotti, et al.^128^11141.340Italy & Australiaindividualistnon-clinicoffendersStateSTAXI-2/State ANGAcceptanceDERS/Nonacceptance-0.27Velotti, et al.^128^11141.340Italy & Australiaindividualistnon-clinicoffendersTraitSTAXI-2/Trait ANGAcceptanceDERS/Nonacceptance-0.15Velotti, et al.^128^111-0Italy & Australiaindividualistnon-clinicoffendersStateSTAXI-2/State ANGSuppressionSTAXI/ANG IN0.23Velotti, et al.^128^111-0Italy & Australiaindividualistnon-clinicoffendersTraitSTAXI-2/Trait ANGSuppressionSTAXI/ANG IN0.34Wang, et al.^129^46420.7565.1Chinacollectivistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGRuminationARS0.70Weber, et al.^130^9323.668Germanyindividualistnon-clinicnon-offendersTraitSTAXI/Trait ANGReappraisalERQ-R-0.20Weindl, et al.^131^22057.940Austriaindividualistnon-clinicnon-offendersTraitSTAXI-2/Trait ANGRuminationDAQ/Anger rumination0.57Wu, et al.^132^35819.1863.12Chinacollectivistnon-clinicnon-offendersTraitSTAXI/Trait ANGRuminationRRQ-R0.28Yavuz, et al.^133^3431.60Turkeycollectivistclinicnon-offendersTraitSTAXI/Trait ANGAvoidanceAAQ-II0.54AAQ-II, Acceptance and Action Questionnaire – version 2; AEQ, Achievement Emotions Questionnaire; ANG, anger; ANG IN, Anger In; BPAQ, Buss Perry Aggression Questionnaire; CAAS, Competitive Anger and Aggressiveness Scale; CECS, Courtauld Emotional Control Scale; CERQ, Cognitive Emotion Regulation Questionnaire; CoSS, Compass of Shame Scale; CWSS, The Coping with Stress Scale; DAR, Dimensions of Anger Reactions; DAQ, Displaced Aggression Questionnaire; DEI, Discrete Emotions Inventory; DERS, Difficulties in Emotion Regulation Scale; DES-IV, Differential Emotions Scale version IV; DRS, Dissipation-Rumination Scale; ECQ-R, Emotion Control Questionnaire – Rumination subscale; ERQ, Emotion Regulation Questionnaire (R, Reappraisal; S, Suppression); ERSCS, Emotion Regulation During Sport Competition Scale; KIMS, Kentucky Inventory of Mindfulness Scale; MAI, Multidimensional Anger Test; ML-STAXI, State Trait Anger Expression and Control Inventory; NAS, Novaco Anger Scale; PAS, Parent Anger Scale; POMS, Profile of Mood Scale; PROMIS, Patient-Reported Outcomes Measurement Information System; PPQ, Present Personality Questionnaire; RIOS, Rumination about Interpersonal Offense Scale; RRQ, Rumination-Reflection Questionnaire; RRS, Rumination Response Scale; SARI, Sadness and Anger Rumination Inventory; SECS, Self-Expression and Control Scale; STAXI-2, State-Trait Anger Expression Inventory, 2nd Edition; TBI-CareQQL, Traumatic Brain Injury Caregiver Quality of Life; Tehran MAS, The Tehran Multidimensional Anger Scale
As shown in Table 2, anger was positively associated with avoidance, rumination, and suppression, and negatively associated with acceptance, and reappraisal. Heterogeneity was significant and high in all analyses.Table 2Meta-analytic associations between anger and emotion regulation strategies.Strategyk (# effect sizes)r95% CIZ**p(Z)Q(df)p(Q)I^2^Acceptance9 (12)-0.32-0.41; -0.23-6.49 < 0.001101.03(11) < 0.00189.11Avoidance12 (12)0.190.08; 0.303.430.00178.88(11) < 0.00186.05Distraction1 (1)------Reappraisal12 (12)-0.13-0.17; -0.05-3.98 < 0.00138.82(11) < 0.00171.67Rumination32 (37)0.420.38; 0.4716.23 < 0.001351.82(36) < 0.00189.76Suppression35 (41)0.240.17; 0.306.89 < 0.001792.56(40) < 0.00194.95CI, confidence interval; df, degrees of freedom.
The methodological quality of most studies (k = 56) was assessed as “poor” (see Supplementary Tables 1–2). Twenty-three studies were rated “fair”, and only 2 were rated “good”. Most studies did not provide sample size justification, and did not control relevant confounding variables. Another common limitation was the cross-sectional design.
Table 3 shows the results of all subgroup analyses.Table 3The results of subgroup analyses on potential moderators.Outcome variableModerator variablek**r95% CI*Z**p(Z)*Q(df)P(Q)AvoidanceType of angerTrait anger-------State anger related to a specific event-----State anger not related to a specific event-----Clinical statusClinical diagnosis30.24[0.04; 0.43]2.390.01724.32(1)0.001Clinical anger-----Non-clinical80.12[0.00; 0.23]2.080.037Criminal statusOffenders-------Non-offenders-----Type of cultureIndividualist80.22[0.07; 0.35]2.870.0040.05(1)0.813Collectivist40.16[-0.04; 0.35]1.580.113Type of designCross-sectional-------Longitudinal-----Study qualityFair40.25[0.12; 0.37]3.72 < 0.00112.93(1)0.008Poor80.16[0.03; 0.29]2.400.016AcceptanceType of angerTrait anger-------State anger related to a specific event-----State anger not related to a specific event-----Clinical statusClinical diagnosis-------Clinical anger-----Non-clinical-----Criminal statusOffenders6-0.35[-0.41; -0.29]-10.06 < 0.00120.24(1) < 0.001Non-offenders6-0.30[-0.44; -0.14]-3.71 < 0.001Type of cultureIndividualist-------Collectivist-----Type of designCross-sectional-------Longitudinal-----Study qualityFair5-0.37[-0.46; -0.27]-7.04 < 0.00113.88(1) < 0.001Poor7-0.29[-0.43; -0.13]-3.52 < 0.001ReappraisalType of angerTrait anger7-0.13[-0.21; -0.04]-2.950.0030.99(1)0.607State anger related to a specific event-----State anger not related to a specific event3-0.09[-0.20; 0.01]-1.7280.084Clinical statusClinical diagnosis-------Clinical anger-----Non-clinical-----Criminal statusOffenders-------Non-offenders-----Type of cultureIndividualist7-0.13[-0.22; -0.05]-3.100.0020.02(1)0.883Collectivist5-0.12[-0.23; -0.00]-2.090.036Type of designCross-sectional-------Longitudinal-----Study qualityFair-------Poor-----RuminationType of angerTrait anger280.42[0.36; 0.47]13.60 < 0.0012.96(2)0.228State anger related to a specific event60.44[0.34; 0.53]7.87 < 0.001State anger not related to a specific event30.40[0.21; 0.57]3.86 < 0.001Clinical statusClinical diagnosis-------Clinical anger-----Non-clinical-----Criminal statusOffenders-------Non-offenders-----Type of cultureIndividualist280.41[0.36; 0.46]13.80 < 0.00130.87(1) < 0.001Collectivist90.45[0.36; 0.53]9.04 < 0.001Type of designCross-sectional-------Longitudinal-----Study qualityFair50.47[0.37; 0.56]8.40 < 0.00132.61(1) < 0.001Poor310.42[0.37; 0.47]14.34 < 0.001SuppressionType of angerTrait anger280.27[0.20; 0.34]7.76 < 0.001297.52(2) < .001State anger related to a specific event70.13[0.06; 0.20]3.78 < 0.001State anger not related to a specific event60.13[-0.05; 0.32]1.380.165Clinical statusClinical diagnosis30.21[0.13; 0.29]5.23 < 0.001136.95(2) < 0.001Clinical anger50.45[0.41; 0.49]20.28 < 0.001Non-clinical330.18[0.17; 0.20]23.29 < 0.001Criminal statusOffenders-------Non-offenders-----Type of cultureIndividualist320.27[0.25; 0.29]29.55 < 0.001106.29(1) < 0.001Collectivist90.11[0.08; 0.13]8.79 < 0.001Type of designCross-sectional-------Longitudinal-----Study qualityFair140.27[0.18; 0.35]5.95 < 0.00141.16(1) < 0.001Poor270.22[0.12; 0.31]4.53 < 0.001Note: Line indicates that the analysis could not be run because there were insufficient studies (k ≥ 3 in at least two subgroups). CI, confidence interval; df, degrees of freedom.
Type of anger (trait anger vs. state anger related to a specific event vs. state anger not related to a specific event) was a significant moderator of the association between anger and suppression, with a medium positive effect size for trait anger, a low positive effect size for state anger related to a specific event, and a non-significant effect size for state anger not related to a specific event. Type of anger was not a significant moderator of the associations between anger and reappraisal, and rumination, and could not be tested as a moderator for the associations with acceptance and avoidance due to the insufficient number of studies per subgroup (i.e., k < 3).
Clinical status (clinical diagnosis vs. clinical anger vs. non-clinical) was a significant moderator of the association between anger and suppression, with a high positive effect size in samples with clinical anger, a medium positive effect size in samples with a clinical diagnosis, and a small-to-medium positive effect size in non-clinical samples. The positive effect size of the association between anger and avoidance was also larger in clinical compared to non-clinical samples (clinical anger could not be included in this analysis due to k < 3). Clinical status could not be tested as a moderator for the associations between anger and acceptance, reappraisal, and rumination because of the insufficient number of studies per subgroup (i.e., k < 3 in more than 1 subgroup).
Criminal status (offenders vs. non-offenders) was a significant moderator in the relation between anger and acceptance, with a larger negative effect size in offenders than in non-offenders. The subgroup analysis could not be conducted for the rest of the associations given that there were fewer than necessary (i.e., k < 3) studies per subgroup.
Type of culture (individualist vs. collectivist) was a significant moderator of the relation between anger and rumination, with a larger positive effect size in samples from collectivist cultures, compared to samples from individualist cultures. It was also a significant moderator of the relation between anger and suppression, with samples from individualist cultures showing higher positive effect sizes compared to samples from collectivist cultures. The type of culture was not a significant moderator of the relations between anger and avoidance, and anger and reappraisal. The subgroup analysis could not be tested in the relation between anger and acceptance given that all study samples were from individualist cultures.
Type of design (cross-sectional vs. longitudinal) could not be tested as moderator given that there were fewer than necessary (i.e., k < 3) studies per subgroup.
Study quality (fair vs. poor; good-quality studies were not included because k < 3) was a significant moderator in the associations between anger and acceptance, avoidance, rumination, and suppression. In all subgroup analyses, fair-quality studies showed a higher effect size compared to poor-quality studies. The analysis could not be run on reappraisal given that there were too few studies (i.e., k < 3) in one subgroup.
Sex distribution. As shown in Table 4, the percent of women was a significant positive predictor of the effect size of the association between anger and acceptance. The association between percent of women and the effect size was not significant in the relations between anger and avoidance, reappraisal, rumination, and suppression.Table 4Results of meta-regression analyses on the relation of sex distribution and mean age with the effect size of the associations between anger and emotion regulation.Outcome variablePredictor variablekslope β95% CIpAvoidanceSample sex distribution (% women)12-0.00[-0.00; 0.00]0.096Sample mean age120.00[-0.00; 0.01]0.348AcceptanceSample sex distribution (% women)120.00[0.00; 0.00]0.028Sample mean age120.02[-0.00; 0.05]0.054ReappraisalSample sex distribution (% women)110.00[-0.00; 0.00]0.093Sample mean age80.00[-0.00; 0.00]0.068RuminationSample sex distribution (% women)33-0.00[-0.00; 0.00]0.360Sample mean age330.00[-0.00; 0.01]0.195SuppressionSample sex distribution (% women)34-0.00[-0.00; -0.00]0.052Sample mean age340.00[-0.00; 0.00]0.783CI, confidence interval.
Age. Mean age was not a significant moderator of the relations between anger and any of the emotion regulation strategies (Table 4).
The funnel plots (Fig. 2A-E) suggested the distribution of the effect sizes was symmetrical for the relations between anger and reappraisal, rumination, and suppression. For the relations of anger with both acceptance and avoidance, the funnel plots showed some asymmetry to the left.Fig. 2Funnel plots of the effect sizes of the association of anger with (A) avoidance; (B) acceptance; (C) reappraisal; (D) rumination; and (E) suppression.
The trim-and-fill procedure imputed two studies in the analysis of the relation between anger and acceptance, which slightly decreased the effect size (r = -0.27). One study was imputed in analysis on anger and reappraisal, and the effect size decreased slightly (r = -0.11). It also imputed three studies in the analysis of the relation between anger and suppression, which decreased the effect size (r = 0.19). No study was imputed in the analysis on anger and avoidance, and anger and rumination.
Egger’s test also indicated that the distribution of effect sizes was symmetrical in all analyses (all ps ≥ 0.069), except that on the relation between anger and avoidance (p = 0.007).
The present results supported consistent associations between anger and multiple emotion regulation strategies, ranging from small to large effect sizes. In addition, the present results identified multiple study-level differences that explained heterogeneity in previous results. They also highlighted multiple methodological limitations of previous studies and suggest critical issues that should be considered in future work.
Anger was positively associated with avoidance, rumination and suppression. These results are in line with theories which have argued that problematic anger involves exaggerated efforts to avoid or block this emotion^14,15^. They also accord with experimental evidence showing that rumination and suppression increase and prolong anger^25,27^. Anger was also negatively associated with the use of acceptance and reappraisal. Theories of psychopathology view the former strategy as fundamental for adaptive emotion regulation^28^, whereas the latter strategy plays a central role in cognitive-behavioral interventions^18^.
Are the present results trivial or do they offer a much-needed perspective on a large and heterogeneous literature? Despite apparent theoretical agreement, it is noteworthy that previous evidence has not always supported the expected pattern of associations between anger and emotion regulation strategies. For instance, while most studies reported positive associations between anger and both avoidance and suppression, and negative associations between anger and both acceptance and reappraisal, some of the previous studies reported correlations in the opposite direction. Furthermore, the effect sizes of these correlations—including rumination—varied widely, ranging from small to large. Therefore, we believe that the present meta-analysis was necessary and contributes to the field by examining whether the previously reported associations between anger and emotion regulation strategies are consistent across studies.
Another question is whether the pattern of emotion regulation strategies associated with anger in the present meta-analysis converges with the results of previous meta-analytic efforts across emotions. Evidence suggests that emotion regulation may partially differ between emotions. For instance, a study found that in anger compared to sadness, people used more emotion regulation strategies, but the strategies did not completely overlap and their effectiveness also differed^134^. Avoidance was used in anger more than in sadness, and the reverse was true for reappraisal^134^. Furthermore, avoidance was associated with lower regulation effectiveness in anger, but not in sadness^134^. A meta-analysis also supported the view that the efficiency of emotion regulation differed between emotions, being lowest in anger and highest in sadness, among the negative emotions that were analyzed^38^. However, this comparison between emotions was possible only across strategies. The present results extend this perspective by identifying the emotion regulation strategies that are positively (i.e., avoidance, rumination, suppression) and negatively (i.e., acceptance, reappraisal) associated with anger. This pattern seems partially different from the one found across emotions^38^ in that suppression is positively associated with anger (and negatively across emotions), and the effect sizes are larger for anger than across emotions.
Another contribution of the present meta-analysis relates to the identification of study characteristics that were found to consistently influence the relations between anger and emotion regulation. For instance, the association between anger and suppression was stronger in studies assessing trait anger compared to those assessing state anger related to a specific event, while studies assessing state anger without a reference event found no significant association. This supports the view that by focusing on context-independent dimensions of anger, dispositional measures capture more of the association with suppression compared to state anger. An alternative explanation is that the larger effect size reflects the similar focus of both trait anger and habitual suppression measures on dispositional aspects (i.e., how one is in general). Assessing state anger without specifying or controlling for the reference event is not recommended because it may introduce uncontrolled contextual variance.
As expected, we found that the positive relation between anger and both suppression and avoidance was larger in clinical compared to non-clinical samples. This is in line with the dimensional approach to psychopathology, which argues that clinical disorders are at the upper extreme of a continuum of individual differences or symptoms in the population^135^. These results also support Roberton’s model^15^, which has argued that one of the maladaptive styles underlying problematic anger involves emotion over-regulation, that is, an exaggerated use of suppression and avoidance aimed at blocking negative emotion. However, it is noteworthy that the association of anger with suppression was large in samples with clinically suggestive levels of anger (i.e., characterized by high levels of negative consequences), and medium in clinical samples selected for any diagnosis of a mental disorder. The likely explanation is that anger problems are present in multiple mental disorders, being more prominent in some disorders (e.g., oppositional defiant disorder, borderline personality disorder, antisocial personality disorder and bipolar disorders)^136–139^ than in others, and that focusing on individuals with high levels of this transdiagnostic symptom captures more of the association between anger and emotion regulation. One additional question is how many of the individuals with anger problems from previous samples had a mental disorder diagnosis? This is relevant in light of debates surrounding the development of specific anger disorder categories in clinical diagnostic^40^. Unfortunately, only one out of five studies in clinical anger samples reported mental disorder diagnoses. This study^83^ suggested a large overlap between anger problems and mental disorders, with 96% of the sample having been diagnosed with psychopathology. Future studies should report clinical status in order to allow for a more extensive examination of this issue.
The comparison between offenders and non-offenders revealed a stronger negative association between anger and acceptance in offenders. This is in line with the view that anger problems are associated with antisocial behaviors, and that low acceptance of anger may be fundamental in the maladaptive regulation of anger that underlies these behaviors^14,15^. However, it is noteworthy that the effect size was also large in the non-offenders’ group, which underscores the hypothesis that mechanisms other than acceptance may be involved in the relation between anger and antisocial behavior.
The positive association between anger and suppression was larger in samples from individualist compared to collectivist countries. As previously suggested^46^, this may reflect that suppression, which typically involves holding back emotion for the sake of social relationships, is less congruent with the self-focused values promoted in individualist cultures. Therefore, when individuals from these cultures use suppression in situations in which someone is in the way of a personal goal, this may result in more anger. Conversely, perhaps by putting personal goals above social well-fare, individualist cultures could promote more anger, which in turn could entail more suppression given that this emotion can harm social image and social relationship. The present meta-analysis cannot discern whether culture promotes anger through suppression, or suppression through anger. The same applies to the positive association between anger and rumination, which was larger in samples from collectivist compared to individualist cultures.
The proportion of women in the sample was positively associated with the effect size of the relation between anger and acceptance. We see two potential compared to men, women may use acceptance to regulate anger more often, or they may use it more efficiently. Previous studies have generally not supported sex differences in acceptance across emotions^140–142^, whereas the studies on anger acceptance included in the present meta-analysis did not examine sex differences. Therefore, we cannot provide a clear explanation, and it remains possible that this finding is driven by other sample characteristics that we did not control for. However, we caution researchers to consider sex differences in their future work on anger and emotion regulation, as these differences may otherwise confound their analyses.
Finally, for all emotion regulation strategies but suppression, the effect sizes were larger in the subgroup of studies rated as qualitatively better. This suggests that the present effects may have been underestimated due to the inclusion of studies that did not meet methodological rigors such as sample size justification, efforts to control potential confounds, and prospective design.
The present meta-analysis has several limitations, which reflect research gaps in the studies that were included. First, most of previous work is cross-sectional and correlational, which precludes any conclusion on the direction of the effects (i.e., anger drives emotion regulation strategies, vice-versa, or both ways). Given that longitudinal designs can shed light on the temporal relations between anger and emotion regulation, we checked the results of the five studies included in the present meta-analysis. All focused on the prospective influence of anger on emotion regulation, and with one exception, did not investigate the reverse pathway. Li, & Xia^99^ were the only ones to consider bidirectional pathways, and supported reciprocal effects between trait anger and rumination. We urge researchers to increasingly use longitudinal designs in future work and examine bidirectional pathways between anger and emotion regulation. This would also contribute to increasing study quality, which is considerably limited in previous studies, for reasons including the scarcity of longitudinal designs, but also issues related to justifying the sample size and controlling for confounding variables.
Second, there is only one study on anger and distraction^71^, and this study focused on behavioral rather than cognitive forms of distraction. Considering experimental evidence showing that when emotional intensity is high, people prefer distraction over reappraisal, and distraction is more efficient than reappraisal^143^, future studies should investigate this strategy in anger. Furthermore, multiple subgroup analyses could not be run in the present analyses, which highlights specific aspects on which data is still limited.
Third, given the central role of emotion regulation in anger problems, studies should increasingly investigate this issue in clinical populations. Previous data is limited and did not allow us to compare the relations between anger and emotion regulation strategies between mental disorders. Anger problems are transdiagnostic, but they may be more prominent in certain disorders compared to others. Empirical investigations of this issue could crucially support efforts to define clinical anger.
Fourth, there was suggestive evidence that publication bias may have confounded the present analyses. While the results of statistical analyses of publication bias were not consistent, which is often the case and reflects the non-overlapping nature of these tests, all analyses except those on reappraisal and rumination showed some degree of publication bias.
Finally, we underscore the importance of focusing on the mediator role of emotion regulation in the relation between anger and aggressive behavior. There has been long-standing theoretical and clinical interest in this hypothesis, but data remains limited. Prospective correlational and experimental designs are the most appropriate approaches for investigating mediation and should be increasingly used in future work.
The present results support the view that anger is consistently associated with the differential use of multiple emotion regulation strategies. This offers an encouraging perspective on a long-standing hypothesis that has been at the heart of clinical theories and interventions in anger and suggests ways in which basic and clinical work in this area could be improved and extended.
The current meta-analysis followed the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) statement^144^ recommendations. The protocol was pre-registered in PROSPERO (registration number CRD42020214171).
The search was conducted in five bibliographical archives (PubMed; PsycINFO; Scopus; Web of Science; Cochrane Central Register of Controlled Trials), from inception until April 3, 2022. We searched for the following keywords in title and “anger”, “emotion regulation”, “reappraisal”, “suppression”, “distraction”, “rumination”, “acceptance”, “avoidance” (see Supplementary Materials for the complete search string). In addition, we searched for relevant articles in the reference section of previous reviews. The search was restricted to articles in English.
Studies were eligible if they met the following (1) they were original quantitative studies, published in peer-reviewed journals; (2) they were conducted in humans; (3) they reported the relation between anger and one or more of the emotion regulation strategies (i.e., acceptance, reappraisal, avoidance, distraction, rumination, suppression); and (4) they provided sufficient statistical data for effect size estimation. We included both correlational studies and studies that compared groups with different levels of anger on emotion regulation.
The aim of this meta-analysis was to investigate the relationship between anger and the habitual or spontaneous use of emotion regulation strategies. Accordingly, we included studies that employed self-report questionnaires to assess the general use of these strategies, as well as field studies (e.g., ecological momentary assessment) that used self-report items to measure emotion regulation in specific contexts. Because these measures differ from those assessing the instructed use of emotion regulation strategies^39^, we excluded studies in which emotion regulation strategies were manipulated. Studies were also excluded if (1) were conference abstracts, letters, dissertations or book chapters; (2) were reviews or meta-analyses not reporting original data; (3) were case studies or qualitative studies; and (4) used the same sample as another eligible study. Adults were the target population of the present meta-analysis and therefore, we excluded studies on samples with a mean age lower than 18 years.
Anger. We focused on anger, which has been distinguished from hostility, which refers to a pervasive aggressive attitude^17,34,145,146^. In line with the phenomenological perspective^147^, only measures of the subjective experience of anger were considered, and not measures of other aspects, such as physiological or behavioral responses related to anger. We included studies that focused on trait anger, state anger unrelated to a specific event, and state anger associated with a specific context (such as a real-world event or an anger-inducing manipulation, for which we considered only post-induction assessments).
Emotion regulation strategies. The following emotion regulation strategies were included in the (1) avoidance (i.e., reluctance to and refraining oneself from experiencing an emotional event); (2) acceptance (i.e., allowing for emotional experiences to unfold naturally); (3) distraction (i.e., moving attention away from an emotional event); (4) cognitive reappraisal (i.e., changing one’s perspective of an emotional event); (5) rumination (i.e., repetitive thinking about an emotional event or experience); and (6) suppression (i.e., inhibiting the behavioral expression of emotion).
All retrieved articles were uploaded in EndNote, and duplicates were eliminated. The titles and abstracts were independently checked by two researchers in order to identify potentially relevant studies. The full texts of the selected studies were also independently assessed by two researchers, based on inclusion and exclusion criteria.
For each study, the following data were (1) identification information (author, year); (2) sample information (i.e., sample size, mean age of the participants, percentage of female participants per study, clinical status, criminal status); (3) emotion regulation strategy; (4) type of anger (i.e., trait, state); (5) instrument used to assess anger; (6) instrument used to assess emotion regulation; (7) effect size; and (8) research design (i.e., cross-sectional, longitudinal). When the relevant data were not reported, authors were contacted for additional details.
The risk of bias in individual studies was assessed by two independent researchers, using tools provided by the National Heart, Lung, and Blood Institute. The Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies^148^ was used for cross-sectional studies, and the Quality Assessment of Case–Control Studies^148^ for studies in which individuals selected for high and low anger levels were compared. The methodological quality of studies was evaluated in line with the following (1) clear research objective; (2) clear specification of the sample (eligibility criteria, justification for the sample size); (3) reliable and valid instruments; and (4) controlling of potential confounding variables. For each criterion, studies were categorized as “yes”, “no”, “cannot determine” (CD), “not applicable” (NA), or “not reported” (NR). In line with previous work^149,150^, the quality of studies was considered good if more than 11 items were rated “yes”, fair if 6–9 items were rated “yes”, and poor if fewer than 6 items were rated “yes”.
Considering our focus on the association between anger and emotion regulation strategies, the Pearson correlation coefficient (r) was used as the index of effect size. In studies comparing individuals selected for anger, we used the group mean, standard deviation and sample size for independent groups to compute Pearson’s r^151^. Following the recommendations of Roth et al. (2018), we chose not to estimate the effect size based on beta coefficients. The analysis was conducted in Comprehensive Meta-analysis version 4.0. The effect sizes were first estimated as Fisher’s z^152^ and then transformed back into the original r-metric for interpretation and presentation^153^. The magnitude of the effect sizes was interpreted in accordance with Gignac and Szodorai^154^: small when r = 0.10, medium when r = 0.20, large when r = 0.30.
In light of the diversity of samples in the included studies, a random-effect model was thus employed in all analyses^155,156^. In order to examine whether there is significant true variance (i.e., beyond random error) in effect sizes between studies, we assessed heterogeneity with the Q statistic test^157^, and estimated the proportion of true variance using the I^2^. An I^2^ index value of 0% indicates no observed heterogeneity, a value of 25% indicates low heterogeneity, of 50%, moderate heterogeneity, and of 75% or above, high heterogeneity^158^.
To examine potential categorical moderators that might explain the true variance in effect sizes across studies, we performed subgroup analysis. The method involves dividing studies into subgroups based on a categorical moderator and comparing the effect sizes between subgroups using tests for heterogeneity (e.g., Cochrahn’s Q) to determine if the differences are statistically significant. We examined the following categorical (1) type of trait anger vs. state anger related to a specific event vs. state anger not related to a specific event; (2) clinical participants with a clinical diagnosis (i.e., any mental disorder diagnosed by a clinician using structured clinical interview) vs. participants with clinically-relevant anger problems (i.e., who scored in the upper quartile on the Trait Anger Scale, TAS > 22, and reported having severe anger problems) but no clinical disorder diagnosis, vs. nonclinical participants; (3) criminal offenders (i.e., participants who were imprisoned or perpetrated intimate partner violence, based on self- or other-reports) vs. non-offenders; (4) type of collectivist vs. individualist, classified as such using the Hofstede’s Country Comparison Tool (https://www.hofstede-insights.com/country-comparison-tool, accessed November 3, 2023); (5) study cross-sectional vs. longitudinal; and (6) study good vs. fair vs. poor. In line with common practice in meta-analysis, a subgroup analysis was run only if there were at least three studies in at least two subgroups. For studies reporting effect sizes for multiple moderator categories on the same sample, in order to ensure that the assumption of independence was not violated in subgroup analyses, we selected only one effect size, specifically the one that added to the less represented subgroup.
Participants’ demographic characteristics (i.e., percentage of women; mean age) and risk of bias across studies were considered potential continuous moderators. A restricted maximum likelihood model for the meta-regression analyses was performed to examine a potential relation between these quantitative study characteristics and the effect size.
Publication bias was examined in three (1) visual inspection of the funnel plot; (b) the trim-and-fill method, which imputes missing studies that would render the funnel plot symmetrical, and estimates an adjusted effect size^153,159^; and (c) the Egger test, which assesses whether the funnel plot is significantly asymmetrical^160^.
Supplementary Information.