This is an open access article distributed under the terms of the Decades after a conflict’s resolution, divergent interpretations of past violence continue to shape political alignments, moral judgments, and calls for justice (Bilali & Ross, 2012; Devine-Wright, 2003; Kurtiş et al., 2018; Noor, Brown, Gonzalez, et al., 2008; Noor, Brown, & Prentice, 2008; Vollhardt, 2012).
This study examines how members of Lebanon’s religious communities interpret the civil war, using Social Identity Theory (Tajfel & Turner, 1986) to explain how group belonging shapes collective memories – shared and socially transmitted interpretations of the group’s past – and how these frames shape construals of victimhood and responsibility (Sahdra & Ross, 2007). Lebanon provides an understudied, multi-factional context marked by overlapping religious and national identities, shifting alliances, external interventions, and postwar amnesty. By examining social identity processes during the Lebanese Civil War, this study answers calls for culturally grounded intergroup research beyond Western contexts (Bilali, 2025; Van Bavel et al., 2016).
Social Identity and Construals of Violence
Social Identity Theory (Tajfel & Turner, 1986) holds that individuals derive part of their self-concept from group memberships, fostering a sense of belonging and self-esteem while motivating favorable ingroup evaluations (Aberson et al., 2000; Brown, 2000; Castano et al., 2002). These processes shape how historical events are remembered and interpreted within collective narratives, particularly in post-conflict contexts where the past is morally contested, and the ingroup is often portrayed as a rightful victim and the outgroup as a perpetrator (Gray & Wegner, 2009). Such identity-driven evaluations inform shared understandings of “what happened” and “who did what.” However, identity-protective distortions do not emerge consistently across all perpetrator contexts (Kazarovytska & Imhoff, 2023).
Self-Categorization Theory (Turner et al., 1987) explains how individuals shift between subordinate (e.g., religious) and superordinate (e.g., national) identities. Subordinate identities accentuate intergroup differences and ingroup suffering, whereas superordinate identities promote recognition of shared experiences and more reconciliatory understandings of collective violence (Čehajić et al., 2008; Gaertner & Dovidio, 2000; Licata et al., 2012). These dynamics shape two key constructs: victimhood beliefs and intergroup responsibility attribution, which are distinct but complementary, especially in civil wars where victim and perpetrator roles overlap.
Collective Victimhood Beliefs
We use collective victimization to refer to processes and experiences of group-based harm, and collective victimhood beliefs to refer to comparative, identity-relevant judgments about who is a ‘victim’ (ingroup vs. outgroup). Recent theoretical developments increasingly use the term “collective victimization beliefs” to emphasize the dynamic, subjective construal of group-based harm (e.g., Vollhardt et al., 2026). In the present research, however, we focus more narrowly on comparative evaluations of victim status between former adversary groups—specifically, whether participants perceive their ingroup as more or less victimized than the outgroup. We therefore retain the term “collective victimhood beliefs” to reflect this evaluative and identity-relevant dimension of post-conflict memory.
Groups may adopt inclusive victim consciousness, recognizing shared suffering (Vollhardt, 2012), or competitive victimhood (CV), claiming greater suffering than others and sometimes denying outgroup harm (Noor, Brown, Gonzalez, et al., 2008; Noor, Brown, & Prentice, 2008). CV serves identity needs: it protects perpetrators’ moral image, reinforces victims’ moral worth, and justifies social or political claims (Bilali & Vollhardt, 2019; Nadler & Shnabel, 2015; Shnabel et al., 2009).
Group members often draw on and reconstruct collective memories of conflict in identity-serving ways, which can sustain CV over time (Mukherjee et al., 2018; Rimé et al., 2015). When perceived ingroup victimization is central to personal identity, individuals are more likely to deny guilt, make exonerating attributions, and oppose conciliatory policies, thereby hindering reconciliation (Jeong et al., 2023; Noor et al., 2012). Such patterns reflect broader ingroup-favoring moral evaluation in intergroup contexts (Bagci et al., 2018; Bilali, 2014; Bilali et al., 2012).
In symmetrical conflicts, victim identities can obscure ingroup wrongdoing (Bilewicz & Stefaniak, 2013; Hirschberger et al., 2016). However, groups may also view themselves as both victims and perpetrators, reflecting the moral complexity of intergroup violence (SimanTov-Nachlieli & Shnabel, 2014). Promoting shared victim or shared perpetrator identities can reduce CV and foster reconciliation (Shnabel et al., 2013).
Building on this logic, identification shapes CV. Identification with subordinate groups (religious/ethnic) heightens CV by reinforcing boundaries, whereas identification with superordinate (national) identities promotes more inclusive interpretations. For instance, Albanians in Kosovo who identified with a shared regional identity (“inhabitants of Kosovo”) reported lower CV (Andrighetto et al., 2012). Competitive victimhood predicts lower levels of forgiveness, trust, and compromise, whereas inclusive construals foster reconciliation (Noor, Brown, & Prentice, 2008; Vollhardt, 2012). Despite Lebanon’s layered religious and national identities, no study has yet examined these dynamics in this context.
We expect participants to perceive greater ingroup than outgroup victimhood (H1a), with this tendency increasing with religious identification (H1b) and decreasing with national identification (H1c). Overall, identification shapes how people construe collective violence, expressed here in competitive victimhood beliefs and intergroup responsibility attributions.
Responsibility Attribution
Beyond shaping victimhood beliefs, identity processes also shape how groups assign responsibility for violence. In intergroup conflicts, groups typically minimize or externalize their own culpability while accentuating others’ (Bandura, 1999; Bilali et al., 2019; Roccas et al., 2006; Rothschild et al., 2013). These attributions reflect motivated efforts to preserve a positive moral image and maintain group cohesion under moral threat, with denial and blame-shifting serving as key defensive strategies (Leidner & Castano, 2012; Lickel et al., 2006; Roccas et al., 2006; White & Branscombe, 2019). Collective memory reinforces these dynamics by foregrounding ingroup suffering and downplaying ingroup agency in harm-doing (Rimé et al., 2015). When victimhood is central to identity, individuals are especially likely to deny guilt and offer exonerating accounts of ingroup behavior (Jeong et al., 2023).
Blaming third parties further protects identity and cohesion, particularly in multi-actor conflicts (Bilali, 2013). In Lebanon, responsibility is often attributed to external actors such as Syria or Israel (Haugbolle, 2010; Licata et al., 2012). Evidence supports this logic: highly identified Turkish participants attributed more responsibility for the Armenian genocide to outgroups and third parties (Bilali et al., 2012), and similar effects emerged in Holocaust remembrance in Hungary (Hirschberger et al., 2016). Strong subgroup identification tends to heighten defensive attributions, whereas superordinate national identification can reduce them and promote reconciliation (Čehajić et al., 2008). Accordingly, we expect participants to attribute more responsibility to the outgroup and third parties than to their ingroup (H2a). This pattern should increase with stronger religious identification (H2b) and decrease with stronger national identification (H2c).
The Case of Lebanon
The Lebanese Civil War offers a context in which overlapping religious and national identities make identity-based interpretations of violence particularly salient. Lebanon officially recognizes 18 sects and is predominantly Muslim (65%) and Christian (35%) (Central Intelligence Agency, 2022). Although rooted in political and socioeconomic inequalities rather than purely religious divisions, the war (1975–1990) crystallized around the Christian–Muslim divide and evolved into a multi-actor conflict shaped by shifting domestic alliances and foreign interventions (Corm, 1986; Salibi, 1988). External actors such as Syria and Israel remain salient in public discourse and are frequently treated as a coherent third-party category (Licata et al., 2012).
Before the war, Christians (particularly Maronites) held disproportionate political power under the 1943 National Pact. During the conflict, power shifted as Muslim factions gained military strength, and later, new asymmetries emerged with Hezbollah’s rise and continued foreign influence (Abul-Husn, 1998; Corm, 1986; Harik, 1996). Although both sides engaged in violence, the conflict was never truly symmetrical; relative power and privilege varied across periods and factions, shaped by domestic alliances and foreign interventions. All major sectarian factions both suffered large-scale victimization and engaged in violence. No group can be meaningfully classified as purely a victim or a perpetrator. This ambivalence makes Lebanon especially suitable for examining how groups construe ingroup victimhood relative to others and assign responsibility for harm.
The war’s 15-year span ended with the Taif Agreement, followed by a general amnesty that blurred distinctions between victims and aggressors under the ethos of “no victor, no vanquished.” While intended to foster reconciliation, the amnesty prevented the emergence of a unified national narrative, leaving community-specific memories that continue to fuel sectarian tensions (Haugbolle, 2010). These fragmented narratives make Lebanon a compelling context for studying how people construe past violence.
Overview of the Present Research
We conducted two studies to examine how members of former adversary groups construe violence and assign victim status and responsibility, focusing on the roles of national and religious identification.
Although Lebanon recognizes 18 sects, political life is organized around two higher-order confessional blocs: Christian (e.g., Maronite, Greek Orthodox) and Muslim (e.g., Sunni, Shia, Druze). Despite their theological and political differences, these groups formed wartime alliances and remain institutionalized in the postwar power-sharing system (Abul-Husn, 1998; Corm, 1986; Salibi, 1988). We therefore operationalized group membership at this broader level. Building on Licata et al. (2012), who focused solely on one denomination (Maronite Christians), our research aimed to compare the two religious groups to provide a more comprehensive understanding of identity-based construals of violence in Lebanon.
Both studies adopted a narrative-based approach, presenting brief, structured accounts of past conflicts. Such narratives mirror how collective memories are socially transmitted and are known to activate identity-based moral reasoning (László, 2013; Liu & Hilton, 2005). In our studies, we used standardized narratives to model these memory frames experimentally, allowing us to examine how identity shapes judgments of victimhood and responsibility under identical informational conditions. Prior research shows that emphasizing ingroup suffering reduces perceived culpability, whereas highlighting ingroup harm-doing elicits defensiveness (Wohl et al., 2006). For example, Bilali (2012) found that participants judged their group’s violence as more justifiable when reminded of its past suffering, underscoring that moral reactions depend on whether the ingroup is portrayed as victim or perpetrator. Accordingly, Study 1 integrated ingroup suffering and harm-doing within single narratives, whereas Study 2 presented them separately, allowing us to test how narrative structure shapes moral judgment and responsibility attribution.
By integrating collective memory and social identity frameworks within a multi-actor conflict, this research extends theories of collective victimhood and responsibility attribution to a complex real-world setting. Together, the studies offer a contextually grounded analysis of how identity shapes interpretations of violence in post-conflict societies.
As researchers, including two Lebanese from different religious communities, we are mindful of how our positionality shapes this work and have engaged in ongoing reflexivity to ensure an inclusive interpretation.
Summary of Hypotheses
Hypothesis 1 (Collective Victimhood): Participants will report greater ingroup victimhood than outgroup victimhood (H1a). This tendency is expected to intensify with stronger religious identification (H1b) and diminish with stronger national identification (H1c).
Hypothesis 2 (Responsibility Attribution): Participants will attribute more responsibility for the war to the outgroup and to third parties than to their ingroup (H2a). This pattern should intensify with stronger religious identification (H2b) and diminish with stronger national identification (H2c).
Study 1
The first study aimed to provide participants with a comprehensive overview of the Lebanese Civil War, covering key events from 1975 to 1990. This broad exposure was designed to assess whether an ingroup-favoring pattern in responsibility attributions and victimhood beliefs could emerge without focusing on specific acts of intergroup violence. In the absence of an official war narrative, we crafted a balanced summary based on the International Center for Transitional Justice's report (2013), reflecting common educational and media representations of the War in Lebanon to ensure ecological validity. The narrative was divided into eight chronological texts (average length = 116 words each) to balance coverage across the 15 years and maintain participant engagement. This segmentation also enabled the inclusion of questions after each text to assess additional variables not analyzed in the present study. Each text described major events initiated by Muslim or Christian factions (see Supplementary Materials). The materials were piloted among Lebanese adults familiar with the conflict to ensure clarity and neutrality, resulting in minor adjustments to wording and tone. After reading these texts, participants evaluated ingroup and outgroup victimhood and responsibility in a single assessment.
Method
Participants
Our study surveyed 405 Lebanese residents, including 218 Christians (127 Maronite, 51 Greek Orthodox, and 40 other Christian denominations) and 187 Muslims (65 Sunni, 92 Shia, 26 Druze, and four other Muslim affiliations). Participants ranged in age from 15 to 79 years, with a mean of 39.59 years (SD = 15.44) and a gender distribution of 56% women and 44% men1. Geographically, 35.8% resided in Beirut, 38.5% in northern Lebanon, and 25.7% in southern Lebanon. Educational background showed that 85.9% had at least one university degree, and 5.4% were teenagers still in school. Additionally, 40.5% had a family member who was a former fighter, and 50.9% had a family member who was a war victim. The average perceived socioeconomic status was 5.60 (SD = 1.39) on a 1-10 scale. Interest in the civil war period was high (M = 8.22; SD = 2.73), as was belief in its ongoing impact on Lebanon today (M = 9.48; SD = 1.09)2.
Procedure
The study used an online survey distributed through social media, targeting voluntary participants who met three criteria: being at least 15 years old, being of Lebanese nationality, and currently living in Lebanon. Participants who did not meet these criteria or who did not complete the questionnaire were excluded. All items were mandatory, so participants who discontinued were not retained. Consequently, the final dataset contained no missing data. The questionnaire was available in English, French, and Arabic to accommodate linguistic diversity, and the translation-back-translation method ensured accuracy across versions. Participants first engaged with the history of the Civil War through eight texts, then assessed group responsibility and victimhood on a subsequent web page. This was followed by completing ingroup identification scales on another page. Data collection occurred from February to April 20193, a period marked by relative peace in Lebanon, free of significant political disturbances.
Measures
All items and measures in the study were assessed using 10-point scales anchored at 1 = “Not at all” and 10 = “Yes, absolutely.”4 The order of the questions followed the order of the main variables presented below.
National Identification
Five items adapted from the collective identification scale developed by Brown et al. (1986) assessed participants’ attachment to Lebanon (e.g., “I often think of myself as Lebanese”; α = .90).
Religious Identification
This variable was assessed using items analogous to those in the national identification scale (5 items), substituting "Lebanon" for "religious community" (e.g., “I feel a strong attachment to my religious community”; α = .93). The two ingroup identification scales were presented on the same web page, with the national identification scale appearing first.
Responsibility Attribution
Participants first rated the general responsibility of Lebanese Christians and Muslims for the civil war. Based on participants’ religious affiliation, these ratings were recoded as ingroup or outgroup responsibility. Responsibility assessments covered both the initiation of the war and conduct during it (e.g., “Christians, in general, are responsible for triggering the civil war”; “Christians, in general, are responsible for some atrocities of the civil war”). Scores for war initiation and atrocities were averaged to create composite measures of ingroup and outgroup responsibility. These components were significantly correlated for ingroup responsibility (Christian sample: r = .512, p < .001; Muslim sample: r = .683, p < .001) and outgroup responsibility (Christian sample: r = .572, p < .001; Muslim sample: r = .651, p < .001).
Participants then rated the responsibility of various third-party actors involved in the conflict (Syria, Israel, Palestinians, Arab countries, European countries, the USA, and other countries). Although these actors differ substantially in their historical roles and geopolitical interests, aggregating them is psychologically meaningful insofar as they represent a common category of external actors intervening in the conflict. From participants’ perspective, these actors are often construed at a similar level of abstraction as “third parties” external to the Lebanese ingroup–outgroup divide. Responsibility ratings for these actors were therefore aggregated into a single third-party responsibility score (α = .87; 14 items). To assess the robustness of this aggregation, we conducted supplementary actor-specific analyses and an exploratory factor analysis; details are reported in the Supplementary Materials.
Collective Victimhood
Participants assessed perceived victimhood by indicating the extent to which Lebanese Christians and Muslims in general could be considered victims of the civil war, using a single item for each group. Participants responded to the question "Which religious denominations can be considered as victim(s) of the civil war?" by rating "Christians in general" and "Muslims in general" on a 1-to-10 scale. Based on respondents’ religious affiliation, these ratings were classified as ingroup or outgroup victimhood variables.
The questionnaire ended with demographic questions.
Statistical Analyses
To test the hypotheses, we conducted repeated-measures mixed ANCOVAs using SPSS (Version 29). For the victimhood analyses, Victimhood (ingroup vs. outgroup) was entered as a within-subjects factor, and group membership (Christian vs. Muslim) as a between-subjects factor. National and religious identification were included as continuous covariates. The model tested the main effect of victimhood and its interactions with group membership and both identification measures. Estimated marginal means were compared using Bonferroni-adjusted pairwise comparisons, and partial eta-squared () was reported as an index of effect size.5 The same analytical strategy was applied to the responsibility attribution variables, with Responsibility (ingroup vs. outgroup vs. third parties) defined as the within-subjects factor. A sensitivity analysis conducted in G*Power indicated that, given the sample size (N = 405), α = .05, and power (1 – β) = .80, the smallest detectable effect size was f = 0.07 ( ≈ .005). This indicates that the study was sufficiently sensitive to detect very small effects of the within–between interaction.
Results
Table 1 presents the means and standard deviations of the main study variables and the zero-order correlations for both groups. A mixed ANOVA indicated stronger identification with the Lebanese nation than with religious groups, F(1, 403) = 350.27, p < .001, = .47, with no interaction with religious group membership, F(1, 403) = 0.001, p = .970, = .00.
Table 1
Descriptive Statistics and Correlations for Christians and Muslims (Study 1)
| Variable | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|---|---|---|---|---|---|---|
| M (SD) | 7.94 (2.39) | 4.87 (3.17) | 6.70 (2.76) | 6.46 (2.97) | 4.67 (2.73) | 5.91 (2.73) | 6.65 (1.82) |
| 1. National identification | .296*** | .090 | .061 | -.093 | .086 | .181* | |
| 2. Religious identification | .200** | .100 | -.083 | -.226** | .164* | .095 | |
| 3. Ingroup victimhood | .036 | .112 | .830*** | .189* | .095 | .307*** | |
| 4. Outgroup victimhood | -.016 | -.158* | .510*** | .287*** | -.052 | .266*** | |
| 5. Ingroup responsibility | -.041 | -.249*** | -.133* | .189** | .391*** | .305*** | |
| 6. Outgroup responsibility | .048 | .154* | .019 | -.022 | .420*** | .189** | |
| 7. Third parties responsibility | .083 | .130 | .043 | .087 | .221** | .297*** | |
| M (SD) | 8.68 (1.94) | 5.63 (2.98) | 7.97 (2.31) | 6.51 (2.81) | 4.31 (2.38) | 6.21 (2.49) | 6.94 (1.65) |
| t test | t(357.47) = 3.40*** d = .34 | t(403) = 2.46* d = .25 | t(363.632) = 4.96*** d = .50 | t(403) = .18 d = .02 | t(403) = -1.44 d = .14 | t(403) = 1.17 d = .12 | t(403) = 1.72 d = .17 |
Note. N = 405. Results for Christian participants (N = 218) are provided in the lower part of the correlation table. Results for Muslim participants (N = 187) are provided in the upper part. Independent-samples t-tests compare mean differences between Christian and Muslim participants.
*p < .05. **p < .01. ***p < .001.
Victimhood Beliefs
Results from the repeated-measures ANCOVA (see Table S1 in the Supplementary Materials for full ANCOVA results) revealed no significant main effect of target group victimhood, F(1, 397) = 0.00, p = .997, = .00, indicating that participants attributed comparable levels of victimhood to the ingroup and the outgroup. None of the interactions involving target group victimhood were significant (all Fs < 1.63, ps > .20), suggesting that this pattern did not vary as a function of religious group, national identification, or religious identification. These findings do not support our first hypothesis (H1a-c).
Responsibility Attribution
The mixed ANCOVA analysis revealed no main effect of target group responsibility, F(2, 794) = 0.33, p = .722, = .00, and no interaction with group membership. However, the Responsibility × Religious Identification interaction was significant, F(2, 794) = 3.11, p = .045, = .02, whereas the Responsibility × National Identification interaction was not, F(2, 794) = 1.54, p = .214, = .01. Bonferroni-adjusted pairwise comparisons showed that, among participants low in religious identification (−1 SD), responsibility attribution increased modestly from the ingroup to the outgroup (p = .050) and more strongly from the ingroup and outgroup to third parties (both ps < .001). Among participants high in religious identification (+1 SD), this gradient was accentuated, with significantly greater responsibility attributed to the outgroup and third parties than to the ingroup (both ps < .001), and a marginal difference between the outgroup and third parties (p = .069). As shown in Figure 1, the expected ingroup-favoring pattern (H2a) emerged only among participants high in religious identification, consistent with H2b (not H2a). National identification did not moderate this effect, providing no support for H2c6.
Figure 1
Estimated Marginal Means of Responsibility Attribution Across Targets (Ingroup, Outgroup, Third Parties) as a Function of Religious Identification (±1 SD), Controlled for National Identification
Note. Error bars represent 95% confidence intervals (Study 1).
Discussion
Study 1 provided partial support for our hypotheses. Although participants did not display a clear ingroup-favoring pattern in victimhood beliefs (suggesting comparable recognition of groups’ suffering rather than competitive victimhood), high religious identifiers attributed greater responsibility to the outgroup and to third parties than to their own group, consistent with previous research linking strong ingroup identification to defensive attributions (e.g., Bandura, 1999; Bilali et al., 2012; Licata et al., 2012). By contrast, low religious identifiers attributed more responsibility to third parties than to either the ingroup or the outgroup. This consistent tendency to blame external actors reflects Lebanon’s long-standing narrative of external interference. Overall, the findings indicate that although religious identity still shapes moral evaluations of the past, there is also evidence of mutual acknowledgment of suffering that may signal early steps toward inclusive victimhood.
However, this study has several limitations. The mixed presentation of events, depicting both ingroup suffering and harm-doing, may have reduced the specificity of ratings of victimhood and responsibility. Moreover, placing the identification scales after the historical narratives and on the same page may have influenced responses through priming or comparison effects. These limitations are addressed in Study 2.
Study 2
Study 2 built on Study 1 by investigating whether distinct or similar outcomes would emerge when participants were presented with clear instances of their ingroup's past victimization or harm-doing. Conducted between April and July 20207, a period marked by COVID-19 lockdowns and ongoing economic and political turmoil in Lebanon, this study diverged from the first by focusing on specific civil war events. Instead of a comprehensive overview, participants were exposed to two distinct texts: one detailing an event of ingroup harm-doing and another of ingroup victimization. The chosen events for this study were the Black Saturday massacre in 1975, attributed to Christian factions, and the Damour village massacre in 1976, attributed to Muslim factions (see Supplementary Materials). These events are well-known among the Lebanese but are interpreted differently across religious groups. The materials were piloted among Lebanese adults familiar with the conflict to ensure clarity and neutrality, resulting in minor adjustments to wording and tone. The design allowed us to isolate the effect of event framing (victim vs. perpetrator) by selecting two incidents that were similar in timing and casualties but differed in the victim–perpetrator dimension. This enabled a focused test of how identification shapes responses to morally affirming versus morally threatening narratives.
Building on Bilali (2012), we expected victimization narratives to be more readily accepted because they align with positive ingroup images, whereas perpetration narratives would trigger defensiveness and justification (Klar & Baram, 2016). Accordingly, participants should show greater ingroup victimhood in response to victimization events and greater ingroup responsibility in response to harm-doing events.
Thus, Study 2 reexamined H1a-c and H2a-c under event-specific conditions and tested whether identification moderates responses to morally threatening narratives. Accordingly, these expectations extend the main hypotheses:
H1d: High religious (or low national) identifiers, in contrast with low religious or high national identifiers, will show more ingroup than outgroup victimhood, especially when reminded of victimization events, as these narratives align with their pre-existing beliefs. This ingroup-favoring pattern is also expected in harm-doing events, with minimized differences in victimhood if acknowledging outgroup victimhood is unavoidable.
H2d: High religious (or low national) identifiers, in contrast with low religious or high national identifiers, will attribute more responsibility to outgroups and third parties than to their ingroup, particularly when reminded of ingroup harm-doing events. If acknowledging ingroup responsibility is unavoidable (as the text explicitly depicts ingroup responsibility in this condition), they will minimize the difference in responsibility between ingroups and outgroups, using symmetrical framing to manage image threats.
Method
Participants
Study 2 involved 338 Lebanese participants residing in Lebanon, comprising 147 Christians (86 Maronite, 28 Greek Orthodox, 17 Greek Catholics, and 16 other Christians) and 191 Muslims (92 Sunni, 72 Shia, 25 Druze, and two other Muslims). The mean age was 44.54 years (SD = 14.02), with participants ranging from 15 to 79 years. The gender distribution was evenly split, with 50% women and 50% men. Geographically, 26.3% lived in Beirut, 41.1% in the north, and 32.5% in the south of Lebanon. A significant majority (80.5%) held at least one university degree, and 2.1% were still in school. Regarding family involvement in the war, 32.2% had a member who actively participated, and 53.3% had a family member who was a victim. The average subjective socioeconomic status was 4.98 (SD = 1.92) on a 1-to-10 scale, and a high frequency of intergroup contact was reported (M = 8.82; SD = 2.16). Most participants (83.1%) reported parents from the same religious community, 14.5% from the same broader religious group but different communities, and 2.4% from different religions.
Procedure
Study 2's data collection mirrored Study 1, using an online survey disseminated through social media in three languages. Key procedural changes included participants first completing ingroup identification scales before reviewing narratives of their group's actions. The scales were presented on separate pages, and the order of presentation was randomized. Each historical narrative and its questions were also presented on individual pages, with the order randomized. Unlike in Study 1, texts remained visible while participants evaluated responsibility and victimhood. Additionally, an interruption was introduced by directing participants to demographic questions between the identification scales and narrative exposure, serving as a buffer to minimize the impact of completing the scales on the assessment of events.
Measures
As in the first study, all items and measures were rated using 1–10-point scales anchored at 1 = “Not at all” and 10 = “Yes, absolutely.” The survey began with the ingroup identification measures, followed by the experimental materials. The order of the two event texts (victimization vs. harm-doing) and their corresponding questions was counterbalanced across participants to control for order effects.
Ingroup Identification
We used the same national (5 items: α = .88) and religious (5 items: α = .87) identification scales as in the first study.
Collective Victimhood
As in Study 1, we assessed each group’s victimhood perception with one item: “To what extent can we consider Christians/Muslims as victims of these events?”. Variables were categorized as ingroup or outgroup based on participants' religious affiliations.
Responsibility Attribution
Responsibility for each group (Christians, Muslims, and third parties) was assessed with a single item: “To what extent can we consider Christians/Muslims/external forces (Arab and non-Arab countries) as responsible for these events?”. Responsibility attributions to Christians and Muslims were categorized as ingroup or outgroup based on participants' religion, as in Study 1. However, unlike Study 1, which combined multiple items to assess third-party responsibility, this study used a single item to assess responsibility for external forces.
Results
Tables 2 and 3 display means and standard deviations for the main variables and zero-order correlations for both groups after exposure to the ingroup perpetrator event and the ingroup victimization event. A mixed ANOVA again showed stronger identification with the Lebanese nation than with religious groups, F(1, 336) = 166.97, p < .001, = .33, with no interaction with religious group membership, F(1, 336) = 0.00, p = .968, < .01, consistent with Study 1.
Table 2
Descriptive Statistics and Correlations for Christians and Muslims When Exposed to Ingroup Perpetrator Event (Study 2)
| Variable | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|---|---|---|---|---|---|---|
| M (SD) | 6.70 (2.77) | 4.24 (2.93) | 5.40 (2.76) | 6.31 (2.91) | 4.58 (2.60) | 4.53 (2.61) | 7.86 (2.59) |
| 1. National identification | .089 | .057 | -.015 | -.066 | .052 | .041 | |
| 2. Religious identification | .457*** | .139 | -.037 | -.023 | .154* | .038 | |
| 3. Ingroup victimhood | .156 | .198* | .273*** | -.027 | .313*** | .037 | |
| 4. Outgroup victimhood | .077 | -.172* | -.044 | .320*** | -.056 | .168* | |
| 5. Ingroup responsibility | -.203* | -.240** | -.329*** | .387*** | .330*** | .055 | |
| 6. Outgroup responsibility | .087 | .245** | .492*** | -.257** | -.129 | -.021 | |
| 7. Third parties resp. | .295*** | .212* | .208* | -.222** | -.220** | .251** | |
| M (SD) | 6.98 (2.79) | 4.51 (2.80) | 5.30 (2.99) | 6.67 (3.01) | 5.63 (2.97) | 4.35 (2.51) | 7.50 (2.75) |
| t-test | t(336) = .91 d = .10 | t(336) = .83 d = .09 | t(336) = -.31 d = -.03 | t(336) = 1.09 d = .12 | t(291.45) = 3.41*** d = .38 | t(336) = -.62 d = .07 | t(336) = 1.24 d = .14 |
Note. N = 338. Results for Christian participants (N = 147) are provided in the lower part of the correlation table. Results for Muslim participants (N = 191) are provided in the upper part. Independent-samples t-tests compare mean differences between Christian and Muslim participants.
*p < .05. **p < .01. ***p < .001.
Table 3
Descriptive Statistics and Scale Correlations for Christians and Muslims When Exposed to Ingroup Victimization Event (Study 2)
| Variable | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|---|---|---|---|---|---|---|
| M (SD) | 6.70 (2.77) | 4.24 (2.93) | 6.54 (2.98) | 5.07 (2.92) | 3.64 (2.22) | 5.37 (2.80) | 7.78 (2.72) |
| 1. National identification | .089 | -.041 | .060 | .125 | .049 | .054 | |
| 2. Religious identification | .457*** | .165* | -.196** | -.114 | .255*** | .069 | |
| 3. Ingroup victimhood | .290*** | .229** | .294*** | .036 | .455*** | .084 | |
| 4. Outgroup victimhood | -.081 | -.118 | -.060 | .285*** | -.059 | .159* | |
| 5. Ingroup responsibility | -.136 | -.269** | -.295*** | .496*** | .233** | .017 | |
| 6. Outgroup responsibility | .086 | .180* | .392*** | -.296*** | -.070 | .052 | |
| 7. Third parties resp. | .251** | .161 | .230** | .075 | .015 | -.117 | |
| M (SD) | 6.98 (2.79) | 4.51 (2.80) | 8.18 (2.29) | 4.43 (3.04) | 3.68 (2.49) | 6.32 (2.90) | 7.99 (2.56) |
| t-test | t(336) = .91 d = .10 | t(336) = .83 d = .09 | t(336) = 5.72*** d = .61 | t(336) = -1.96 d = .22 | t(336) = .16 d = .02 | t(336) = 3.04** d = .33 | t(336) = .73 d = .08 |
Note. N = 338. Results for Christian participants (N = 147) are provided in the lower part of the correlation table. Results for Muslim participants (N = 191) are provided in the upper part. Independent-samples t-tests compare mean differences between Christian and Muslim participants.
*p < .05. **p < .01. ***p < .001.
Victimhood Beliefs
We conducted a repeated-measures mixed ANCOVA, as in Study 1 (see Table S3 in the Supplementary Materials for full ANCOVA results), this time incorporating event-type exposure (past harm-doing vs. past victimization) as an additional within-subject factor alongside target-group victimhood (ingroup vs. outgroup). A sensitivity analysis conducted in G*Power indicated that with N = 338, α = .05, and power (1 – β) = .80, the smallest detectable effect size was f = 0.064 ( ≈ .004). This indicates that the study was sufficiently powered to detect very small effects for the within–between interactions.
The analysis yielded a marginally significant main effect of victimhood, F(1, 330) = 3.79, p = .052, = .01, suggesting a weak tendency to attribute greater victimhood to the ingroup than to the outgroup (H1a). Consistent with H1b, a significant Victimhood × Religious Identification interaction, F(1, 330) = 10.12, p = .002, = .03, indicated that highly religious participants differentiated more sharply between ingroup and outgroup victimhood (Figure 2) than low identifiers, who perceived both groups as equally victimized. The Victimhood × National Identification interaction was not significant, F(1, 330) = 0.91, p = .340, = .00 (H1c).
Figure 2
Interaction Between Target-Group Victimhood and Religious Identification, Adjusted for National Identification (Study 2)
Event type also affected perceived victimhood, F(1, 330) = 20.78, p < .001, = .06: participants attributed more ingroup victimhood following ingroup victimization than after ingroup harm-doing events, Fs(1, 330) = 140.41, p < .001, = .30, and F(1, 330) = 21.74, p < .001, = .06, respectively. A significant Victimhood × Event × Religious Identification interaction, F(1, 330) = 6.15, p = .014, = .02, indicated that religious identification amplified event-based differences: high identifiers showed a stronger ingroup-favoring pattern when confronted with ingroup victimization than low identifiers (Figure 3). When exposed to ingroup wrongdoing, high identifiers attributed similar levels of victimhood to both groups, whereas low identifiers acknowledged more outgroup suffering.
Figure 3
Interaction Between Target of Victimhood, Event Type, and Religious Identification (±1 SD), Controlling for National Identification (Study 2)
Finally, a four-way interaction with national identification, F(1, 330) = 6.80, p = .010, = .02, (Figure 4) indicated that the strongest ingroup-favoring pattern in victimization contexts emerged among participants high in both religious and national identification. In contrast, acknowledgment of outgroup suffering was greatest among those low in religious identification (regardless of national identification) when confronted with ingroup wrongdoing. No significant interactions with group membership emerged.
Figure 4
Interaction Between Event, Target of Victimhood, Religious and National Identification (Study 2)
Overall, results partially supported Hypothesis 1: the competitive victimhood pattern was marginally present (H1a), appeared mainly among highly religious participants (H1b), and was particularly pronounced when confronted with past ingroup victimization (H1d), but not among those with low national identification (no support for H1c).
Responsibility Attributions
A repeated-measures ANCOVA with target group responsibility (ingroup, outgroup, third parties) revealed a significant main effect, F(1.84, 608.10) = 6.13, p = .003, = .02, confirming clear differentiation among targets. Consistent with H2a, participants attributed more responsibility to third parties than to the outgroup (p < .001) and more to the outgroup than to the ingroup (p < .001).
A significant Responsibility × Religious Identification interaction (H2b), F(1.84, 608.10) = 3.15, p = .048, = .01, showed that high religious identifiers attributed less responsibility to their ingroup and more to the outgroup and third parties. In contrast, low identifiers showed smaller differences (see Figure 5). The Responsibility × National Identification interaction, F(1.84, 608.10) = 1.72, p = .182, = .01, was not significant (no support for H2c).
Figure 5
Estimated Marginal Means of Responsibility Attribution Toward the Ingroup, Outgroup, and Third Parties as a Function of Religious Identification (Low vs. High), Controlled for National Identification (Study 2)
The Event × Responsibility interaction indicated event-based variation, F(1.80, 593.69) = 16.82, p < .001, = .05: after harm-doing events, responsibility ranked third parties > ingroup > outgroup (ps < .001 for all pairwise comparisons); after victimization, it ranked third parties > outgroup > ingroup (ps < .001 for third parties vs. others; p = .003 for outgroup vs. ingroup).
A significant Responsibility × Event × National Identification interaction, F(1.80, 593.69) = 6.11, p = .003, = .02 (Figure 6), further clarified the previous effects. Responsibility-attribution patterns were similar across low and high national identifiers in victimization events. However, in harm-doing events, low national identifiers attributed more responsibility to the ingroup, whereas high national identifiers minimized it, assigning similar levels of responsibility to the ingroup and outgroup.
Figure 6
Estimated Marginal Means of Responsibility Attribution Toward the Ingroup, Outgroup, and Third Parties as a Function of National Identification (Low vs. High) and Event Type (Ingroup Harm-Doing vs. Ingroup Victimization), Adjusted for Religious Identification (Study 2)
Finally, the significant four-way interaction (Figure 7) provided additional insight, F(1.80, 593.69) = 4.42, p = .015, = .01. When facing ingroup harm-doing, only participants with low levels of both religious and national identification attributed more responsibility to the ingroup than to the outgroup. In the three other identification configurations, ingroup and outgroup responsibility levels were similar. When exposed to ingroup victimization, participants high in religious identification, regardless of national identification level, attributed substantially more responsibility to the outgroup than to the ingroup, whereas attribution patterns among low religious identifiers varied as a function of national identification. Third parties consistently received the highest blame, regardless of event type and identification levels.
Figure 7
Estimated Marginal Means of Responsibility Attribution (1–10 Scale) as a Function of National and Religious Identification (±1 SD) and Event Type (Ingroup Harm-Doing vs. Ingroup Victimization) (Study 2)
Overall, results partially supported H2: the ingroup-favoring pattern in responsibility attribution (H2a) was stronger among high religious identifiers (H2b), while national identification did not reduce ingroup-outgroup differences (no support for H2c). Evidence for H2d was limited: religious identification primarily structured responsibility attribution patterns, with no additional differentiation by national identification among high religious identifiers. National identification differentiated attribution patterns only among low religious identifiers, with low religious–low national participants showing a reversal of ingroup–outgroup responsibility across harm-doing and victimization conditions. No significant interactions with group membership emerged.
Discussion
Study 2 extended Study 1 by presenting participants with specific civil war events depicting their ingroup as either a victim or a perpetrator, allowing us to examine how event framing and identification jointly shape interpretations of past violence.
Unlike Study 1, which found no ingroup-outgroup differences in victimhood, Study 2 revealed clear ingroup-favoring victimhood, particularly following an ingroup victimization event. This pattern was strongest among highly religious identifiers, consistent with expectations and supporting the role of religious identification in reinforcing competitive victimhood beliefs (Noor, Brown, & Prentice, 2008). Unexpectedly, these effects were amplified among participants high in both religious and national identification, rather than among those high in religious but low in national identification.
Responsibility attributions showed a similar overall pattern: participants assigned more responsibility to outgroups and third parties than to their ingroup, especially when religious identification was strong, as in Study 1. The event type further qualified this pattern. When the ingroup was portrayed as a perpetrator, participants low in both religious and national identification attributed more responsibility to the ingroup than to the outgroup. In contrast, highly religious participants, regardless of national identification level, displayed more symmetrical responsibility attributions between groups. This partially supports H2d, suggesting that religious identification heightened defensiveness primarily by minimizing ingroup–outgroup differentiation rather than by directly displacing blame. Contrary to expectations, strong national identification did not attenuate this defensiveness, and no effects of group membership emerged.
Together, these results highlight how identity salience and event framing jointly shape moral interpretations of conflict, reinforcing collective defensiveness in post-conflict memory.
General Discussion
This research investigated how Lebanese Christians and Muslims interpret their country’s civil war through victimhood beliefs and responsibility attributions, and how these vary with religious and national identification. Based on Social Identity Theory (Tajfel & Turner, 1986) and work on identity-based motivated reasoning and moral defensiveness (Bilali & Vollhardt, 2019; Roccas et al., 2006), we expected ingroup-favoring patterns—greater perceived ingroup victimhood and reduced ingroup responsibility—especially among strong religious and weak national identifiers.
To our knowledge, this is the first systematic investigation of victimhood beliefs and responsibility attribution in Lebanon and among the few to integrate third-party blame into models of post-conflict collective memory (Bilali et al., 2012; Bilali, 2013; Çelebi et al., 2014; Licata et al., 2012). Whereas earlier research examined clearer perpetrator-victim asymmetries, our study extends this literature to a multi-actor conflict marked by overlapping roles, advancing understanding of how collective memory and identity interact when moral boundaries are ambiguous.
Across two studies using narrative reminders of the war, one broad (Study 1) and two event-specific narratives (ingroup harm-doing vs. ingroup victimization in Study 2), support for the hypotheses was mixed. Despite being conducted at different political moments (2019, during relative stability, vs. 2020, amid an economic and health crisis), several patterns were consistent. Participants attributed more responsibility to outgroups, especially third parties, than to their ingroup, and these patterns were strongest among highly religious identifiers. Notably, these patterns did not differ significantly between Christian and Muslim participants, indicating that identification strength and narrative framing mattered more than categorical group membership.
Victimhood Beliefs
Victimhood effects emerged primarily in Study 2. When reminded of their group’s past suffering, participants attributed greater ingroup than outgroup victimhood, particularly those high in religious identification. This suggests that strong religious identifiers sought to defend their group’s moral image when confronted with threatening narratives, consistent with identity-driven moral reasoning and selective memory (Bandura, 1999; Baumeister & Hastings, 1997; Bilali et al., 2012; Leidner & Castano, 2012; Noor, Brown, & Prentice, 2008; Noor et al., 2012; Rothschild et al., 2013; Sahdra & Ross, 2007).
Contrary to expectations, national identification did not offset the defensiveness associated with religious identification. Participants high in religious identification showed similar patterns of ingroup victimhood regardless of their level of national identification. This challenges common ingroup identity models (Gaertner & Dovidio, 2000), which posit that superordinate identities reduce ingroup-favoring patterns, and suggests that in Lebanon, where national identity is historically intertwined with sectarian narratives (Salibi, 1988), superordinate identification may lack the integrative force assumed in other contexts. Positive correlations between national and religious identification nonetheless point to complex identity configurations in internal religious–ethnic conflicts.
Across samples, ingroup and outgroup victimhood were positively correlated, indicating some capacity for inclusive victimhood (Vollhardt & Bilali, 2015). Recognizing shared suffering can foster empathy and reconciliation, suggesting that interventions emphasizing mutual loss may be effective. Research on historical perspective-taking (Bilali & Vollhardt, 2013) shows that acknowledging outgroup suffering can evoke empathy and moral reflection without requiring the denial or minimization of one’s own group’s suffering, an approach well suited to Lebanon’s fragmented memory landscape.
Responsibility Attributions
Responsibility attributions reflected both defensive and integrative tendencies. Across both studies, participants attributed more responsibility to the religious outgroup and, even more, to third parties than to their own group, consistent with identity-protective and moral disengagement strategies (Bandura, 1999; Bilali et al., 2012; Roccas et al., 2006). As expected, religious identification amplified this ingroup-favoring tendency (Study 1), suggesting that moral defensiveness intensifies with the salience of religious identity (Noor et al., 2012; Rothschild et al., 2013).
In Study 2, event type further qualified these patterns. When reminded of their group’s past victimization, highly religious participants (regardless of national identification level) increased outgroup blame. By contrast, when confronted with ingroup harm-doing, they shifted toward symmetrical responsibility attributions, minimizing differences between groups. Thus, strong religious identification did not simply increase blame; it promoted flexible defensiveness—externalizing responsibility when identity-congruent and adopting moral equivalence when acknowledging ingroup wrongdoing was unavoidable (Shnabel et al., 2013). Participants low in both religious and national identification, however, were more likely to acknowledge ingroup responsibility under harm-doing conditions, suggesting greater responsiveness to narrative content than to identity concerns (Bilali et al., 2019). Together, these patterns indicate that identification strength shapes not only who is blamed but also the structure of responsibility judgments (symmetrical vs. asymmetrical) in post-conflict contexts, thereby refining models of moral reasoning and responsibility attribution in multi-actor conflicts.
Contrary to expectations, national identification neither mitigated nor uniquely amplified religious defensiveness. Rather, it appeared largely neutral in shaping responsibility attributions among highly religious participants. In the Lebanese context, where national identity has historically been articulated through selective and contested narratives (Salibi, 1988), such patterns suggest that dual identification may coexist with moral symmetry rather than promote accountability. While this symmetrical framing may reduce overt blame and support coexistence, it can also obscure responsibility, highlighting the limits of identity-based integration for fostering deeper reconciliation.
Third-party blame was particularly salient. Across both studies, external actors were consistently assigned the greatest responsibility, surpassing both ingroup and outgroup blame. This mirrors Lebanon’s broader public discourse, which often frames the conflict as externally instigated rather than domestically driven. Our results thus position third-party blame as a double-edged narrative: while such narratives preserve ingroup virtue and can temporarily strengthen national cohesion, they deflect accountability and constrain moral repair (Bilali, 2013; Licata et al., 2012).
The absence of group-membership effects suggests that categorical religious affiliation alone does not structure moral interpretations of the Lebanese civil war. This convergence is further supported by the finding that Christian and Muslim participants did not differ in their relative levels of national versus religious identification, suggesting comparable identity configurations across groups. This may reflect the conflict’s shifting perpetrator–victim roles, which blur moral boundaries across groups, as well as the dominance of identification strength over group category in shaping motivated reasoning. Moreover, decades of post-war silence and shared narratives emphasizing external responsibility may have fostered convergent collective memory frames among Christians and Muslims, reducing divergence in responsibility and victimhood judgments. At the institutional level, Lebanon’s post-war amnesty law exemplifies this dynamic. By attributing the war’s causes to foreign interference, it reduced immediate tensions yet hindered the formation of a shared national narrative and long-term collective memory work (Staub, 2006).
Generalizability and Context-Specificity
Although rooted in Lebanon’s history, the persistence of religious identification as a moral anchor (and the defensiveness it produces) likely generalizes to other post-conflict settings where religion and politics intertwine (e.g., Bosnia, Northern Ireland). The potential for inclusive victimhood and shared responsibility may characterize multi-actor conflicts in which no group holds absolute moral authority. By contrast, the overlap between national and religious identification is context-specific, reflecting Lebanon’s confessional system and uneven representation. A similar pattern has been documented in Northern Ireland, where national and religious identities remain intertwined and selectively inclusive even after political settlements (Muldoon et al., 2007). The prominence of third-party blame, framing domestic conflict as externally driven, may also extend to societies shaped by colonial or proxy-war legacies, where external involvement continues to influence collective memory (Bilali et al., 2012; Čehajić‐Clancy & Bilewicz, 2020).
Limitations and Future Directions
Our convenience sample limits generalizability, and future work should recruit more socioeconomically diverse participants to better capture variation in war memories and identity positions. Moreover, our single-item victimhood measures capture perceived asymmetries but do not fully reflect the multidimensional constructs of competitive versus inclusive victimhood. Our findings should therefore be interpreted as addressing relative judgments of victim status in this context rather than the full breadth of comparative victim belief dimensions. Future research should employ validated multi-item scales. Given the sensitivity of responsibility judgments in polarized settings, self-report measures may underestimate defensiveness. Behavioral, implicit, or experimental methods could detect subtler forms of identity protection. Moreover, the difficulty of isolating victimization and perpetration reflects a real-world feature of intergroup conflict: collective memories are relational and cyclical, shaped by contested narratives and mutual interpretations of harm (Bilali, 2012; Licata et al., 2007). Qualitative or mixed-method designs could better capture how participants themselves construct meaning around such events.
Future research should test whether counter-narratives highlighting outgroup victimization can foster reconciliation. Our results suggest openness to inclusive victimhood (Vollhardt & Bilali, 2015) and that victimhood beliefs are malleable: individuals capable of ingroup-favoring attributions also recognize shared losses. Experimental interventions using testimonials or media could examine whether acknowledging outgroup suffering enhances empathy and reduces defensiveness. Building on historical perspective-taking (Bilali & Vollhardt, 2013) and work on managing competing collective memories (Licata et al., 2007), initiatives in Lebanon, such as dialogue, education, or media, may promote more balanced moral appraisals and readiness for reconciliation. Longitudinal designs should further investigate how identity salience, narrative framing, and moral reasoning interact over time. Integrating identity theory with perspective-taking interventions can clarify how inclusive collective memories emerge and how they may help repair divisions in Lebanon and comparable societies.
Conclusion
Our studies show that religious identification continues to shape moral interpretations of Lebanon’s civil-war past, fostering selective memory and defensive attributions (Baumeister & Hastings, 1997; Noor, Brown, & Prentice, 2008; Rothschild et al., 2013). Although national identification remains widely endorsed, it neither counteracts nor consistently attenuates these ingroup-favoring patterns, suggesting limits to its capacity to serve as a unifying framework. Yet the coexistence of ingroup-favoring patterns with recognition of shared suffering reveals a capacity for inclusive remembrance, in which group members acknowledge shared losses and accountability.
By showing how identity strength and narrative framing jointly shape post-conflict interpretations, this research advances understanding of collective memory in morally ambiguous conflicts. It highlights the enduring influence of religious identity, the limits of national identity as a unifying framework, and the ambivalent role of third-party blame in sustaining or softening intergroup divisions.