Direct and indirect punishment of norm violations in daily life

The study investigates when and how people deploy different punishment strategies in real-world settings. While cooperation can be sustained via direct and indirect reciprocity and partner choice, prior empirical work largely stems from laboratory paradigms that may not capture the relational, situational, and emotional factors shaping punishment in daily life. Punishment can be direct (overt physical or verbal confrontation) or indirect (gossip, social exclusion/avoidance), with direct strategies posing higher retaliation risks but potentially greater immediacy and effectiveness, and indirect strategies carrying lower immediate risk but possibly slower or less direct impact. The authors hypothesize that punishment decisions reflect cost–benefit tradeoffs: punishers will favor direct confrontation when there is more to gain from changing the offender’s behavior (e.g., the offender is highly valued; the punisher is personally victimized) and will favor indirect strategies when retaliation risks are higher (e.g., violations judged more severe; offenders are more powerful). Emotions, particularly anger and disgust, are expected to differentially motivate direct versus indirect punishment. The study aims to document punishment responses to norm violations in daily life across varied relationships and contexts, testing preregistered predictions about relational valuation, victim status, perceived moral wrongness, relative power, and moral emotions.

Theoretical and empirical work indicates multiple mechanisms maintain cooperation, including direct reciprocity, indirect reciprocity, and partner choice. Punishment features prominently across these accounts, with evidence for costly punishment in lab games and across societies, and for indirect strategies like gossip and ostracism promoting cooperation. Field studies suggest low-cost strategies (e.g., withholding benefits) may be more frequent than confrontational punishment. Existing research often focuses on interactions between strangers and controlled tasks (second- and third-party punishment), which may not generalize to everyday relational contexts. Decision rules for punishment likely weigh benefits (valuing the offender, self-victimization) and risks (severity of violation as a cue to offender commitment; offender power). Moral emotions are central motivators of punishment, with sociofunctional accounts differentiating anger (approach, confrontation) and disgust (avoidance, social distancing, coordination of third-party punishment). Prior vignette and survey work suggests victims may favor direct punishment more than third parties, and that disgust relates to indirect punitive tendencies, but real-world longitudinal evidence has been limited.

Design: Preregistered longitudinal experience-sampling study over two weeks with daily mobile surveys and two follow-up surveys (7–14 days post-incident) assessing punishment responses to norm violations in daily life. Materials, data, and code were preregistered and are available on OSF. Sample: N = 257 Dutch adults recruited via two panel agencies (Flycatcher, Link2Trials). Demographics (of 256 respondents): 66.1% female; age M = 39.15 (SD = 16.02; 18–75); 60.5% bachelor’s+; SES ladder M = 6.49 (SD = 1.54); 90.2% born in the Netherlands. Ethics approved by VU Amsterdam (VCWE-2018-052); informed consent obtained. Procedure: Lab intake (questionnaires, demographics, physical formidability, incentivized decision tasks) followed by daily assessments for 14 days (SMS at 19:00 with 6-hour response window; one reminder). Median time to open: 1 h 1 m; median completion: 7 m. Two follow-ups sent after the daily phase (for week 1 and week 2 events, 7–14 day lag; 24 h response window). Follow-up median open: 1 h 20 m; median completion: 6 m. Compensation up to €59. Daily assessments: Each day assessed up to two norm violations: (1) self-relevant (participant victim), (2) other-relevant (observed/learned victimization of someone else). If none, assessed social or non-social events. Overall response rate: 80.27% (2888 daily surveys). In 1236 surveys (42.80%), at least one violation was reported; total violations k = 1468 (self-relevant k = 901; other-relevant k = 567). Physical presence during violation recorded (present k = 968; absent k = 500). Measures: For each violation, participants described the event and offender. Relationship type; offender gender; emotional closeness; valuation of relationship via welfare tradeoff ratio (WTRown: amount foregone €0–10 for offender to receive €10) and perceived offender WTRother; perceived moral wrongness (1–5) and harmfulness (1–5); situational interdependence (subset of Situational Interdependence Scale); situational power (1 = definitely the offender, 5 = definitely myself). For other-relevant violations, analogous measures for the victim. Emotions: Arrays of facial expressions (Radboud Faces Database) to assess anger, disgust, fear, sadness, happiness (1–5), plus general affect valence (1 very negative to 5 very positive). Punishment motivations (5-point Likert): physical confrontation, verbal confrontation, gossip, social exclusion. Punishment behaviors (Yes/No): direct confrontation, gossip (told someone else when offender absent), social avoidance (avoided social contact). Open-ended responses captured behavior details. Follow-ups: For each earlier violation, reassessed closeness, WTRs, emotions, and punishment behaviors on days after the violation, with open-ended descriptions. Analysis: Hierarchical data structure (violations nested within days within subjects). Punishment motivations analyzed via linear mixed models (random intercepts/slopes for days and subjects; autocorrelation structure). Punishment behaviors analyzed with binary logistic GEEs (nesting within days and subjects; autocorrelation). Continuous IVs decomposed into within-person centered and person-average effects; additional random effects for within-person-centered variables and binary IVs. Participant gender controlled in all analyses. All tests two-sided. Data manipulation and analyses in R and SPSS. Data exclusions: none beyond using completed assessments; if duplicate completions occurred, first response retained.

• Prevalence of punishment strategies: • Motivations: Participants reported stronger motivations for gossip (M = 2.82, SD = 1.24) and social exclusion (M = 2.79, SD = 1.37) than for physical (M = 2.13, SD = 1.18) or verbal confrontation (M = 2.45, SD = 1.28); F(3, 2537.27) = 111.19, p < 0.001. • Daily behaviors: Gossip occurred in 44.1% of events, confrontation in 35.4%, avoidance in 34.8%; Wald χ²(2) = 27.64, p < 0.001. When physically present, gossip (45.9%) ≈ confrontation (42.6%), both > avoidance (37.4%; gossip vs avoid: Wald χ²(1) = 12.07, OR = 1.37, p = 0.001; confront vs avoid: Wald χ²(1) = 5.61, OR = 1.31, p = 0.018). • Follow-up behaviors (7–14 days later): Confrontation 24.9%, gossip 45.4%, avoidance 27.9%; Wald χ²(2) = 145.81, p < 0.001; same pattern as daily.
• Benefits of changing offenders’ behavior: • Valuation (WTRown): Higher valuation predicted more direct confrontation (Wald χ²(1) = 38.44, b = 0.16, p < 0.001) and less gossip (Wald χ²(1) = 13.40, b = −0.09, p < 0.001) and less avoidance (Wald χ²(1) = 28.94, b = −0.14, p < 0.001). Motivations decreased overall with higher valuation, especially for exclusion (Table 1). • Self-relevance: Being the victim strongly increased confrontation (Wald χ²(1) = 84.98, b = 1.07, p < 0.001) and made confrontation more likely relative to gossip (Wald χ²(1) = 16.74, OR = 1.86, p < 0.001) and avoidance (Wald χ²(1) = 15.90, OR = 1.92, p < 0.001).
• Risks of retaliation: • Moral wrongness: Greater perceived wrongness associated with more avoidance (Wald χ²(1) = 12.06, b = 0.25, p = 0.001) and more gossip (Wald χ²(1) = 10.22, b = 0.23, p = 0.001), but less confrontation (Wald χ²(1) = 4.48, b = −0.15, p = 0.034). Motivations increased most for exclusion (Table 2). • Relative power: Lower power predicted greater use of gossip (Wald χ²(1) = 5.28, b = −0.17, p = 0.022) and avoidance (Wald χ²(1) = 4.23, b = −0.16, p = 0.040), and less confrontation (Wald χ²(1) = 36.21, b = 0.48, p < 0.001). Motivational patterns mirrored these effects (Table 3).
• Emotions: • Anger: Positively associated with overall punishment intensity (motivations F(1, 372.55) = 67.16, p < 0.001; behaviors Wald χ²(1) = 7.62, p = 0.006) without differential preference for strategy. • Disgust: Differentially associated with indirect strategies; higher disgust increased likelihood of gossip (Wald χ²(1) = 11.43, OR = 1.32, p = 0.001) and avoidance (Wald χ²(2) = 3.98, OR = 1.20, p = 0.046) relative to confrontation; motivational interactions significant (Table 4). Overall, participants used confrontation more when personally victimized, more powerful, and when offenders were valued; they used gossip and avoidance more for severe violations and when they had less power or valued offenders less. Gossip was the most frequent response overall.

Findings demonstrate that punishment of norm violations in daily life is strategic and context-sensitive rather than unconditional. Individuals favor direct confrontation when benefits of adjusting the offender’s behavior are high (e.g., self-victimization, high offender value, higher relative power), and they favor indirect punishment when retaliation risks loom large (e.g., severe/morally wrong violations, lower relative power). Gossip emerged as the most common response overall, while confrontation was equally common when immediate intervention was possible; both exceeded avoidance. Emotional drivers align with sociofunctional theories: anger broadly intensifies punitive responding, whereas disgust selectively promotes indirect strategies that distance the punisher and potentially recruit others. These results enrich evolutionary and normative enforcement models by highlighting heterogeneity in punitive strategies and the roles of valuation, power dynamics, and moral severity in real-world interactions, including differences between second- and third-party-like contexts where direct confrontation is rarer.

The study contributes longitudinal, ecologically valid evidence that people deploy direct and indirect punishment in daily life based on calculated tradeoffs between benefits and retaliation risks, moderated by relational value, victim status, perceived moral wrongness, power, and moral emotions. It shows gossip’s prominence and the conditional use of confrontation, offering actionable insights for theories of cooperation and norm enforcement. Future research should incorporate multiple punishment strategies and contextual moderators (power asymmetries, social networks, offense costliness) into formal models, directly measure the actual costs imposed on offenders by different strategies, and examine how modern platforms (e.g., social media) facilitate indirect punishment and the spread of moral outrage.

• The inclusive operationalization of punishment (e.g., verbal condemnation, reputation manipulation, social exclusion) may include responses that do not always impose effective costs on offenders.
• Daily, prospective assessments reduce recall bias but might heighten participants’ attention to norm violations and increase intervention frequency; timing could also underestimate behaviors that occur after more delay (addressed partly with follow-ups).
• Costs imposed on offenders were not directly measured; effectiveness comparisons among strategies rely on theory and proxies.
• Technical issues reduced delivery of the second follow-up to some participants, potentially affecting follow-up response rates.