Broken Promises: Betrayal and Support for Violence in Intergroup Relations

Abstract

In diverse societies, minority groups may face challenges when events signal exclusion from a superordinate identity (SOI) shared with the relevant majority groups. We examine how such SOI threats relate to hardline political attitudes, focusing on betrayal as a potential mechanism. A cross-sectional study of Ethiopian Jews in Israel (N = 276) showed that priming an SOI threat was associated with support for violent resistance via betrayal. A two-wave study of Arab-Muslims in Israel (N = 165) showed that a real-time SOI-threatening event predicted betrayal and, in turn, increased support for violence, particularly among those with stronger baseline SOI. An additional two-wave study of Israeli Jewish women (N = 584) during the recent Gaza war extended this framework to a broader SOI shared with women worldwide: stronger baseline SOI predicted higher expectations of solidarity, which, when undermined by SOI threat, was associated with greater betrayal and hawkish wartime policy support.

Introduction

Promoting sustainable coexistence in the face of increasing ethno-racial and religious diversity is a paramount challenge in contemporary society. The potential of superordinate identities to foster improved intergroup relations is well recognized: seeing one’s group as part of a broader shared category that includes both the ingroup and the outgroup can reduce intergroup bias (Gaertner & Dovidio, 2000). However, sociopolitical occurrences, such as promises of egalitarian reform that go unfulfilled, can jeopardize social cohesion and undermine the sense of belonging that minority members feel to these overarching identities (Adam-Troian et al., 2020; Saguy et al., 2020; Verkuyten & Yildiz, 2007). Such exclusionary events can serve to disillusion minority members, making them feel (or realize) that the promises embedded in a shared “we” were broken. The present research examines this dynamic and its potentially damaging implications for intergroup relations and minority political engagement.

We address two questions: First, how do minority group members react politically when a superordinate identity (SOI) shared with the majority is threatened? Second, what psychological mechanisms can explain such reactions? We test these questions across three studies, each conducted in a different intergroup context. We researched an ethnic minority in Israel—Ethiopian Jews (Study 1) focusing on a threat to their shared identity with non-Ethiopian Jews in Israel; and a national/religious minority, Arab-Muslim citizens of Israel (hereafter Arab-Muslims; Study 2) ¹ —focusing on threats to their SOI with Jewish Israelis. Study 3 extended this framework to a gender-based global context, examining Israeli Jewish women’s responses to perceived abandonment by international women following the October 7 attacks by Hamas on Israel, and specifically on Israeli women (Botbol, 2023; Gettleman et al., 2023). Across contexts, we examined whether betrayal links SOI threat to hardline and defensive political responses, including support for violence and endorsement of hawkish policy stances during war.

Superordinate Identities Among Minorities

According to the common ingroup identity model (Gaertner & Dovidio, 2000), identification with an overarching category that includes both the ingroup and outgroup can extend ingroup favoritism to the former outgroup. Positive outcomes of such SOI include helping and prosocial support (Dovidio et al., 2010; Kunst et al., 2015), support for cooperative policies (Beaton et al., 2008), and more positive outgroup attitudes and resource allocation (Houlette et al., 2004). These effects are often attributed to increased perceived similarity and connectedness, which can counteract prejudice and bias (Saguy, 2018).

At the same time, research has also highlighted more complex dynamics in unequal settings (Cakal et al., 2011; Dixon et al., 2012; Saguy et al., 2009). For example, in structurally unequal contexts, emphasizing commonality can increase expectations of fairness among low-power groups while reducing their motivation for collective action—without necessarily increasing redistribution by the high-power group (Saguy et al., 2009). Similarly, inducing common humanity increased forgiveness among Indigenous Australians toward colonizers while reducing collective action intentions (Greenaway et al., 2012). Thus, SOIs can foster trust and positive orientations (Wohl & Branscombe, 2005), while also cultivating optimism about a more egalitarian future.

But what happens when these implied promises go unfulfilled? Under such conditions, minority group members may feel disillusioned as expectations of fair treatment embedded in an SOI are violated. We refer to this experience as a threat to SOI. The present research examines the consequences of such threats, particularly those triggered by exclusionary events, and the psychological mechanisms through which they may translate into hardline political attitudes.

SOI Threat and Hardline Political Attitudes

Although SOI threat has rarely been examined in the way we propose, substantial work documents political consequences of perceived exclusion among minorities. Across political psychology and political science, discrimination and exclusionary threats are often linked to political action, including protest participation, civic engagement, and increased in-group mobilization (Besco, 2024; Bilodeau et al., 2023; Chan, 2022; Dai et al., 2023; Oskooii, 2020). For instance, political discrimination increased both mainstream and ethnic-based engagement among Asian and Black minorities in Britain (Oskooii, 2020), and threats to citizenship mobilized Palestinian citizens of Israel by increasing turnout, online discourse, and social-movement participation (Weiss et al., 2023). Broadly, across diverse minority contexts, exclusion is frequently associated with defensive in-group action (Besco, 2024; Tardivo, 2025).

Yet exclusion does not only mobilize civically. Feelings of alienation and neglect among minorities have also been linked to more hardline political attitudes, including support for violent collective action (Adam-Troian et al., 2020) and willingness to join radical groups (Bélanger et al., 2019). For example, perceived oppression predicted radicalism intentions among individuals identifying with Catalonia in the Catalonian-Spanish conflict (Lobato et al., 2020); social alienation legitimized political violence among Maghrebi French minorities (Troian et al., 2019); and relative deprivation among Muslim immigrants predicted support for terrorist attacks and violent intentions (Obaidi et al., 2019). Among Muslims in Germany, perceived discrimination was associated with antisystem attitudes, reflected in increased support for violence and Islamism alongside reduced democratic support (Grewal & Hamid, 2024). Similarly, discrimination predicted endorsement of violence and rejection of representative democracy among Brussels youth (Mansoury Babhoutak et al., 2020). Taken together, exclusion has been shown not only to facilitate political/civic participation but also at times to produce hardening and aggressive, even violent, political expression.

At the same time, exclusion can also reshape how minorities locate themselves relative to the broader society. Work in political science has shown that perceived discrimination was linked to partisan alignment and shifts in political identity, particularly among minorities navigating their place within a national majority (Kuo et al., 2017). Work on the “integration paradox” similarly shows that more acculturated or native-born minority members often report more discrimination and alienation. For example, Swedish-born residents with migrant backgrounds in stigmatized neighborhoods perceived higher group discrimination and, in turn, reported lower national identification and institutional trust (Esaiasson et al., 2024), and highly integrated American Muslims—socioeconomically incorporated and culturally fluent—were especially likely to perceive societal and political discrimination (Lajevardi et al., 2020). These accounts typically interpret such patterns as reflecting elevated expectations of fairness, equal treatment, or democratic inclusion among more integrated minorities. However, these expectations remain largely implicit in the theorizing and are not modeled as relational mechanisms.

In contrast, we explicitly conceptualize SOIs as relational commitments that embed promises of equality, dignity, and mutual concern between minority and majority groups. As such, shared identities generate relationship-based expectations of fairness and solidarity. When an SOI is threatened, this threat may be experienced as a relational violation that signals the shared “we” is no longer being upheld, evoking especially intense disillusionment that not only erodes belonging but also undermines imagined futures of coexistence (Saguy & Sobol-Sarag, 2022). Evidence from the Arab Druze minority in Israel illustrates this dynamic: exclusionary developments, such as the Nationality Bill, predicted decreased identification with the Israeli SOI and increased support for violent collective action, particularly among youth whose military service had emphasized shared identity with Jewish Israelis (Saguy & Sobol-Sarag, 2022; Saguy et al., 2020). These patterns are consistent with the idea that prior inclusion and investment in a shared “we” can heighten the sting of exclusion, yet this work did not assess SOI or expectations before exclusionary events, nor did it identify a specific psychological mechanism linking SOI threats to hardline political attitudes.

Two questions therefore remain. Without observing prior SOI and its implied expectations before exclusionary turns, it is difficult to determine whether SOI threats are especially consequential for those previously invested in a shared “we.” In addition, existing work leaves unclear what affective process translates SOI threat into hardline political expression. We propose that group-level betrayal captures this process: a moral-relational response to broken superordinate promises that can convert disillusionment into hawkish or confrontational political expression.

A Sense of Group-Level Betrayal

Betrayal arises when relationship-relevant norms are violated, breaking implicit social contracts (Elangovan & Shapiro, 1998; Finkel et al., 2002; Stackhouse et al., 2023). Although typically examined in interpersonal relationships, betrayal is also relevant in intergroup contexts. We propose that minority members may experience betrayal when the majority’s harmful actions—or inactions (e.g., silence in the face of discrimination)—are perceived to deviate from shared values believed to govern the relationship. In this way, SOI-threatening events can communicate that promises of fairness and equality were broken, eliciting betrayal.

Prior research on group-level betrayal has largely focused on intragroup contexts (e.g., disloyalty by fellow members), particularly among those high in ingroup identification who expect norm adherence (Masson et al., 2016; O’Brien et al., 2012; Tsordia et al., 2024). While centered on within-group dynamics, this work supports the premise that betrayal rests on expectations derived from a shared identity; expectations that may also guide relations between minorities and majorities when an SOI is salient.

When such expectations are violated, betrayal can motivate punitive responses. Interpersonal work suggests it elicits desires for vengeance (Finkel et al., 2002; Fitness, 2001), and intragroup research shows that norm violators are often punished or excluded (Ditrich et al., 2022; Mendoza et al., 2014). These reactions can reflect attempts to restore control and meaning in the face of vulnerability (Fitness, 2001; Webber et al., 2018). Moreover, betrayal-related responses can extend beyond the original perpetrator: unforgiveness and betrayal can spill over into displaced revenge against broader collectives associated with the offender (Chou et al., 2022; Stackhouse et al., 2023), and threat-related responses can carry over across outgroups (Bouman et al., 2014). Together, this work suggests that when identity-based expectations are violated, efforts to restore meaning and control may take confrontational, defensive, and sometimes redirected forms.

Building on these considerations, we propose that in exclusionary contexts, minority group members who perceive a threat to their shared identity with the majority may experience heightened betrayal, particularly when they previously felt strongly connected to the superordinate group (i.e., stronger baseline SOI). Betrayal, in turn, should be associated with hawkish or hardline political expression, including the promotion of force. This may be expressed as support for violence directed at the perceived betraying majority outgroup, or as hawkish policy preferences that endorse force toward another target in ways that nonetheless reflect distancing from the shared identity. By contrast, when baseline SOI is weak, exclusion may evoke less disillusionment and weaker links between threat and betrayal (Figure 1). Moreover, expectations may function as a key link between superordinate identification and betrayal: stronger identification should foster stronger expectations of fairness and solidarity, which, when violated under threat, give rise to betrayal (Figure 2). Critically, Studies 2–3 assess baseline SOI (and, in Study 3, solidarity expectations) prior to later betrayal and hardline outcomes, enabling prospective tests of the hypothesized pathway.

Figure 1

Theoretical model: Betrayal as a mediator between SOI threat and support for hardline political attitudes, moderated by baseline SOI.

Figure 2.

Theoretical model: Expectations and betrayal as sequential mediators linking baseline SOI to support for hardline political attitudes under threat.

The Current Research

To test these propositions, we conducted three studies with different minority groups in three intergroup contexts. Across studies, we focus on two forms of hardline political attitudes: (a) support for violence directed at the perceived betraying majority outgroup itself (Studies 1–2), and (b) hawkish policy preferences that imply support for force and aggressive measures toward another target in ways that reflect distancing from the shared identity (Study 3). Together, these studies test the moderated pathway depicted in Figure 1 and the expectations-based mechanism depicted in Figure 2.

Study 1 focused on Ethiopian Jews in Israel, an ethnic minority comprising roughly 2% of the population. Despite socioeconomic disadvantage, Ethiopian Jews share a strong religious and historical bond with the Jewish-Israeli majority, anchored in a shared Jewish identity and a collective narrative of return (Getahun & Keynan, 2022). We tested whether perceived threat to this shared identity related to betrayal and, in turn, support for violence against the Jewish-Israeli majority, providing an initial test of the Figure 1 pathway in a context where baseline SOI is presumed relatively strong.

Study 2 examined these dynamics among Arab-Muslim citizens of Israel, a religious minority whose national SOI as “Israeli” is more heterogeneous (Hermann et al., 2019). Using a two-wave design, we assessed SOI at Time 1 during an unprecedented inclusive Jewish–Arab government that emphasized Arab rights and promised greater equality. Eight months later (Time 2) and following the government’s dissolution, we measured perceived SOI threat, betrayal, and support for violence. This design tests whether SOI threat following an exclusionary political turn is most strongly linked to betrayal and violent inclinations among those who previously reported higher superordinate identification (Figure 1).

Study 3 extended the framework to a transnational SOI by examining Jewish-Israeli women’s identification with women worldwide. At Time 1 (post–October 7), we measured SOI and expectations of solidarity from international women as baseline predictors. Approximately 2 months later (Time 2), participants read a brief reminder summarizing an investigative New York Times report on sexual and gender-based violence during the October 7 attacks and, amid a perceived lack of international recognition and solidarity, reported betrayal by international women and indicated hawkish wartime policy preferences. This study tests whether baseline SOI predicts stronger solidarity expectations and, in turn, greater betrayal and more hawkish attitudes (Figure 2).

Across the three studies, we evaluate whether disruptions to a shared “we” are linked to hardline responses via betrayal, testing both when this process is strongest as a function of baseline SOI (Figure 1) and one pathway through which it unfolds via solidarity expectations and betrayal (Figure 2).

Study 1

Study 1 tested whether the association between perceived threats to a shared SOI and support for violent resistance can be explained by feelings of intergroup betrayal among Ethiopian Jews in Israel. We conducted the study a few months after the police shooting and killing of an Ethiopian man—an event that gained much public attention and openly challenged Ethiopian Jews’ sense of shared belonging with the Jewish-Israeli majority (TOI Staff & Agencies, 2019). In a correlational design, we assessed perceived SOI threat (IV), betrayal by non-Ethiopian Jewish Israelis (mediator), and support for violent resistance (DV), controlling for demographics.

Method

Open Science Statement

Data, analysis code, and online supplemental materials (OSM; all studies) are available on OSF: https://osf.io/b2wcf/. Studies 1–3 were not preregistered.

Participants

Participants were recruited through snowball sampling among Ethiopian citizens in Israel. The study was originally planned as an experiment, aiming to manipulate Ethiopian participants’ perceived threat to their SOI with the Jewish-Israeli majority via different framings of the police shooting and killing of Solomon Tekah, an Ethiopian-Israeli young man (TOI Staff & Agencies, 2019). Participants were randomly assigned to a high-threat, low-threat, or control condition. An a priori power analysis (Monte Carlo simulation) targeting the indirect effect (a × b) indicated N = 250 was required for 80% power (β = .21 per path; Richard et al., 2003); we aimed for approximately 100 participants per condition to meet this target.

The initial sample included 311 participants, of whom 33 were excluded for missing data on the outcome variable and 2 were excluded for failing both attention checks. The final sample consisted of 276 participants (M_age = 29.91, SD = 6.84; 63.4% female). To assess whether outcome missingness was systematic, we estimated a logistic regression predicting availability of the support for violent resistance outcome from experimental condition, SOI threat, and covariates; no predictors were significant (see OSM, Table S1). Because the manipulation did not reliably affect the focal variables, we report Study 1 as a correlational test of the proposed associations; details of the manipulation, power analysis with code, and ancillary analyses are provided in the OSM (see Open Science Statement).

Procedure and Measures

Full item wordings are listed in the OSM. Participants first reported demographics: gender, age, education, socioeconomic status (SES). They then completed the following measures:

SOI Threat

Four items assessed perceived exclusion and lack of solidarity from the Jewish-Israeli majority (1 = do not agree at all, 7 = agree to a large extent; for example, “The majority of the Israeli public sees Ethiopians as ‘others’,” “The rest of the Israeli public does not care about the discrimination against Ethiopians”; α = .71).

Intergroup Betrayal

Five items adapted from interpersonal betrayal research (Fitness, 2001) were used to assess betrayal (e.g., “I feel betrayed by Israeli society,” “My trust in Israeli society has been damaged”; α = .88; 1–5 response scale).

Support for Violent Resistance

Support for violent resistance was assessed with one item (“To what extent do you support a violent protest against police violence toward Ethiopians?”; 1 = do not agree at all, 7 = agree to a large extent).

Results

Descriptive statistics and zero-order correlations are shown in Table 1. We tested the hypothesized mediation using PROCESS in R (Model 4; Hayes, 2022), specifying SOI threat as the independent variable, betrayal as the mediator, and support for violent resistance as the dependent variable, controlling for gender, age, education, and SES (Figure 1).

Table 1.

Means, Standard Deviations, and Correlations With Confidence Intervals.

Variable	M	SD	1	2	3	4	5
1. Age	29.91	6.84
2. Education	3.75	1.23	.23*** [0.11, 0.34]
3. SES	2.73	1.36	.22*** [0.10, 0.33]	.21*** [0.09, 0.32]
4. Threat	4.90	1.10	−.12 [−0.23, 0.00]	−.11 [−0.23, 0.01]	.02 [−0.10, 0.14]
5. Betrayal	3.85	1.44	−.11 [−0.23, 0.01]	−.14* [−0.26, −0.03]	−.10 [−0.22, 0.02]	.51*** [0.42, 0.59]
6. Support for Violent Resistance	2.61	2.21	.01 [−0.11, 0.13]	−.21*** [−0.32, −0.09]	−.08 [−0.19, 0.04]	.14* [0.02, 0.26]	.23*** [0.12, 0.34]

Note. Values in square brackets indicate the 95% confidence interval for each correlation.

*

p < .05. ***p < .001.

As predicted, SOI threat was positively associated with intergroup betrayal (b = 0.65, p < .001, 95% CI [0.52, 0.79]). Betrayal predicted greater support for violent resistance controlling for threat (b = 0.31, p < .01 [0.10, 0.51]). SOI threat predicted support for violent resistance in the total-effect model (b = 0.23, p < .05 [0.03, 0.50]) but not after including betrayal (b = 0.07, p = .62 [−0.20, 0.34]). The indirect effect via betrayal was significant (indirect effect = 0.20; 5,000 bootstrap samples; 95% bootstrapped CI [0.05, 0.38]). Complete model estimates are reported in the OSM. We also conducted a mediation sensitivity analysis (Imai et al., 2010); the indirect effect would be reduced to zero at ρ = .20 (OSM, Tables S6–S7).

Discussion

Study 1 provided initial evidence that intergroup betrayal may function as a mechanism linking perceived threats to a shared SOI with support for violent resistance. Consistent with work showing that betrayal can elicit retaliatory motives (Finkel et al., 2002), betrayal at the intergroup level was associated with heightened minority support for violence toward a majority group perceived as having betrayed them.

Study 1 was nevertheless cross-sectional and did not include an assessment of baseline SOI or expectations prior to the threatening event. Thus, although Ethiopian Jews are often characterized as having relatively strong identification with the Jewish-Israeli majority (Getahun & Keynan, 2022; Silbereisen & Brenick, 2012), we could not test whether SOI threat is especially consequential among those higher in prior SOI. Studies 2 and 3 address these limitations by incorporating baseline measures and longitudinal designs.

Study 2

Study 2 used a two-wave design to assess baseline SOI prior to a threatening event. Participants were Arab-Muslim citizens of Israel. Time 1 data were collected shortly after the formation of an inclusive Jewish–Arab government that publicly promised societal change in favor of Arab rights. We reasoned that such signals of shared interests could strengthen some Arab-Muslim citizens’ sense of commonality with the Jewish-Israeli majority. We anticipated that the coalition’s political fragility could make it short-lived and that a subsequent exclusionary turn might provide an opportunity to assess SOI threat as it unfolded. Time 2 data were collected after the collapse of this government (which lasted 15 months), which we expected would signal undermined prospects for inclusion.

We tested whether perceived SOI threat following the government’s collapse would be associated with heightened betrayal and greater support for violence, especially among participants who previously reported stronger SOI. Thus, Study 2 tested whether baseline SOI moderates the indirect effect of SOI threat on support for violence via betrayal (Figure 1).

Method

Participants

Participants were recruited through a survey company specializing in the Arab population in Israel. Data were collected at two time points: February 2022 (Time 1; shortly after the formation of the first Israeli government to include an Arab party) and September 2022 (Time 2; 3 months after the government’s dissolution and 2 months before national elections). The initial sample included 402 Arab-Muslim participants who completed the Time 1 phone survey. At Time 2, 165 participants (41% retention) completed the survey again (M_age = 51.84, SD = 15.69; 31.5% women). An a priori power analysis (Monte Carlo simulation) targeting the index of moderated mediation (a₂ × b) indicated N = 210 was required for 80% power (effect sizes from Study 1, Saguy et al., 2020, and Richard et al., 2003). Our final N = 165 retains 80% power to detect moderate effects (e.g., a₂ = 0.25, b = 0.25; see OSM for details, code, and Figures S4–S5).

Procedure and Measures

SOI threat (T2) served as the independent variable, baseline SOI (T1) as the moderator, intergroup betrayal (T2) as the mediator, and support for violence (T2) as the dependent variable (see Figure 1). Baseline support for violence (T1) was included as a covariate, alongside demographic variables assessed at T2. Full item wordings are provided in the OSM.

Background Variables

Participants reported gender, age, education, and SES; these were assessed at T2 and included as covariates.

Superordinate Identity

Baseline SOI (T1) reflects strength of identification with the SOI and was assessed with a thermometer item measuring identification with the Israeli identity (0 = does not belong at all, 100 = very much belongs).

SOI Threat

SOI threat (T2) was measured with the same four-items used in Study 1, adapted by replacing references to “Ethiopian” with “Muslim” citizens (α = .72).

Intergroup Betrayal

Betrayal (T2) was assessed using the same five items from Study 1 (α = .85; 1–5 response scale).

Support for Violence

Support for violence (T2) was assessed with two items: the Study 1 item (assessing support for violent protest against police violence) and an additional item assessing support for violent protest more generally in the Arab–Jewish context. Responses (1–7) were averaged to form a composite (r = .81, p < .001).

Results

Descriptive statistics and zero-order correlations are shown in Table 2. Attrition analyses indicated small retention differences by age and gender: participants retained at Time 2 were slightly older on average (M_T1 = 46.82, SD_T1 = 16.26; M_T2 = 50.68, SD_T2 = 15.27; p = .016) and more likely to be male (χ²[1] = 4.62, p = .032), whereas baseline SOI, SOI threat, betrayal, support for violence, SES, and education did not differ by retention status (see Tables S8–S9 in the Online Supplement). In a logistic regression predicting Time 2 completion from Time 1 variables (Table S10), age remained significant (OR = 1.02, 95% CI [1.00, 1.03], p = .013), whereas gender was reduced to a marginal association (Female vs. Male: OR = 0.67 [0.44, 1.02], p = .063). We included age and gender as covariates in all analyses (alongside demographic variables).

Table 2.

Means, Standard Deviations, and Correlations With Confidence Intervals.

Variable	M	SD	1	2	3	4	5	6	7	8
1. Age	51.84	15.69
2. Education	2.89	1.74	−.21** [−0.35, −0.06]
3. SES	1.92	1.26	−.25** [−0.39, −0.10]	.40*** [0.26, 0.52]	.
4. SOI Threat T2	4.88	1.41	−.15 [−0.29, 0.00]	.13 [−0.02, 0.28]	.16* [0.00, 0.30]
5. Betrayal T2	3.06	1.15	−.10 [−0.25, 0.06]	.08 [−0.07, 0.23]	−.02 [−0.18, 0.13]	.44*** [0.30, 0.55]
6. SOI T1	47.35	35.36	.10 [−0.06, 0.25]	−.13 [−0.27, 0.03]	−.20** [−0.35, −0.05]	−.25** [−0.39, −0.10]	−.39*** [−0.51, −0.26]
7. Support for Violence T1	2.05	1.68	−.07 [−0.22, 0.08]	−.07 [−0.22, 0.08]	−.01 [−0.16, 0.14]	−.11 [−0.26, 0.04]	.12 [−0.04, 0.26]	−.13 [−0.28, 0.02]	−.13 [−0.31, 0.05]
8. Support for Violence T2	2.14	1.80	−.11 [−0.26, 0.04]	−.07 [−0.22, 0.08]	.16* [0.01, 0.31]	.11 [−0.04, 0.26]	.31*** [0.17, 0.45]	−.15 [−0.29, 0.01]	−.01 [−0.20, 0.17]	.28*** [0.14, 0.42]

Note. Values in square brackets indicate the 95% confidence interval for each correlation. SOI = superordinate identity.

*

p < .05. **p < .01. ***p < .001.

The primary goal of Study 2 was to test whether the indirect effect of SOI threat on support for violence via betrayal varies as a function of baseline SOI. We conducted a moderated mediation analysis using PROCESS in R (Model 7; Hayes, 2022). We specified SOI threat (T2) as the predictor, betrayal (T2) as the mediator, and support for violence (T2) as the outcome, controlling for baseline support for violence (T1). We further tested whether baseline SOI (T1) moderates the association between SOI threat (T2) and betrayal (T2; Figure 1).

Moderation Analysis

As predicted, the SOI threat × baseline SOI interaction significantly predicted betrayal (b = 0.004, SE = 0.002, p < .01). Simple slopes showed that SOI threat was positively associated with betrayal at mean baseline SOI (b = 0.29, SE = 0.06, p < .001) and at high baseline SOI (+1 SD: b = 0.44, SE = 0.08, p < .001), but did not reach significance at low baseline SOI (−1 SD: b = 0.15, SE = 0.08, p = .054). Thus, the association between SOI threat and betrayal was strongest among participants higher in baseline SOI (Figure 3).

Figure 3.

Conditional effects of SOI threat on betrayal.

Conditional Indirect Effects

Replicating the basic mediation pattern, SOI threat (T2) predicted higher support for violence at T2 (controlling for T1 levels) indirectly via betrayal (T2; indirect effect = 0.19, 95% bootstrapped CI [0.08, 0.31]). The moderated mediation model indicated that this indirect effect differed across levels of baseline SOI: the index of moderated mediation was significant (Index = 0.003, bootstrapped SE = 0.002, 95% bootstrapped CI [0.0002, 0.006]). Consistent with the theoretical model, the indirect effect was strongest at high SOI (+1 SD; b = 0.30, SE = 0.11, 95% CI [0.12, 0.55]), smaller but significant at mean SOI (b = 0.21, SE = .08, 95% CI [0.08, 0.40]), and not significant at low SOI (−1 SD; b = 0.12, bootstrapped SE = 0.08, 95% bootstrapped CI [−0.01, 0.31]). Complete statistical estimates are provided in the OSM. ² We also conducted a mediation sensitivity analysis for the conditional indirect effect at high baseline SOI (+1 SD); the indirect effect would be reduced to zero at ρ = .30 (Table 3; OSM, Tables S15–S16).

Table 3.

Moderated Mediation: Dependent Variable Model.

Predictor	B (SE)	95% CI	Β	t	p
(Intercept)	0.91 (0.88)	[−0.83, 2.64]	—	1.03	.303
Gender: Female	−0.37 (0.27)	[−0.91, 0.17]	−.10	−1.35	.179
Age	−0.00 (0.01)	[−0.02, 0.01]	−.04	−0.55	.582
Education	−0.19 (0.08)	[−0.35, −0.03]	−.18	−2.33	.021*
SES	0.33 (0.11)	[0.10, 0.55]	.23	2.89	.004**
Support for Violence T1	0.24 (0.08)	[0.09, 0.40]	.23	3.16	.002**
SOI Threat T2	−0.03 (0.10)	[−0.23, 0.17]	−.02	−0.29	.769
Betrayal T2	0.50 (0.13)	[0.25, 0.74]	.32	3.94	< .001***
R 2	.226
Observations	165

Note. Values in square brackets indicate the 95% confidence interval for each B. B = unstandardized regression coefficient; β = standardized coefficient; SOI = superordinate identity.

*

p < .05. **p < .01. ***p < .001.

Discussion

Study 2 corroborated Study 1 and strengthened confidence in the proposed model by incorporating baseline SOI. Betrayal mediated the association between SOI threat and support for violence, and this indirect effect was strongest among participants who reported higher baseline SOI. Thus, when individuals who previously felt more connected to the superordinate group perceived that shared identity as threatened, they were more likely to experience betrayal, which in turn related to greater support for violence. When baseline SOI was low, the indirect association via betrayal was weaker and not significant.

Study 3

Study 3 examined whether the proposed process extends beyond ethnic–religious minorities to a cross-national SOI and whether it generalizes to a different form of hardline political expression. We also introduced expectations of solidarity as a proximal pathway-linking baseline SOI to later betrayal. Jewish-Israeli women were conceptualized as a minority subgroup within a broader global community of women. We theorized that stronger identification with women worldwide at Time 1 would predict stronger expectations of solidarity, which—if later perceived as violated—would give rise to greater betrayal and, in turn, stronger endorsement of hawkish wartime policies at Time 2 (Figure 2).

Data were collected in the months following the October 7 attacks during the war in Gaza, amid public attention to reports of sexual violence committed during the attacks and debates about perceived delayed or insufficient recognition by some international women’s rights actors and UN-affiliated entities (Botbol, 2023; United Nations, Office of the Special Representative of the Secretary-General on Sexual Violence in Conflict, 2024; UN Women, 2023).

The study was originally designed as an experiment. At Time 1, we assessed SOI with women worldwide and expectations of solidarity without mentioning October 7 or international reactions. At Time 2, all participants read a brief summary adapted from an actual New York Times investigative report describing reports of widespread sexual violence during the October 7 attacks (Gettleman et al., 2023). Participants were then randomly assigned to information suggesting that women worldwide were likely to believe the report (low threat/high solidarity), or to doubt it (high threat/low solidarity). A third group of participants received no information about belief/disbelief (control; see OSM).

The manipulation did not affect any of the variables in the study. This is in line with Study 1 and suggests that SOI threat may be difficult to induce via brief experimental messages (see also Hopkins et al., 2020; we return to this point in the Discussion). Given contemporaneous debates about recognition and solidarity, the shared report exposure prior to random assignment may have made SOI threat salient across conditions, yielding elevated betrayal regardless of subsequent information. We therefore opted to treat Study 3 as a longitudinal correlational test of the sequential mediation model, retaining experimental condition as a covariate.

Method

Participants

Participants were recruited via the Israeli online panel Panel4All. Data were collected in two waves during the Israel–Gaza war: March 2024 (T1) and May 2024 (T2). At Time 1, 816 participants began the survey. After excluding respondents without consent (n = 2), duplicate entries (n = 2), and cases with more than 20% missing responses (n = 47), the final T1 sample comprised 765 Jewish-Israeli women aged 18 and above. At Time 2, 631 participants returned. After excluding cases without valid consent (n = 7), duplicate entries due to repeated participation (n = 17), and cases with excessive missing outcome data (n = 20), the final longitudinal sample consisted of 584 participants (76.3% retention from T1). Attrition was not systematically related to T1 study variables or demographics (all ps ≥ .142; see OSM, Tables S19–S20).

Demographics (age, education, income) and ideological covariates (political ideology, religiosity, feminist ideology) were included as controls.

An a priori power analysis (Monte Carlo simulation) targeting the sequential indirect effect (a × b × c) indicated N = 290 was required for 80% power (β = .21 per path; Richard et al., 2003). Our final N = 584 substantially exceeded this requirement (see OSM for details and code).

Procedure

At Time 1 (March 2024), participants completed measures of SOI with, and expectations of solidarity from women worldwide, along with background variables (age, education, income) and ideological covariates (political ideology, religiosity, feminist ideology). At Time 2 (May 2024), participants read a brief news-style summary adapted from a New York Times article describing reports of sexual violence during the October 7 attacks (Gettleman et al., 2023). Subsequently, participants were randomly assigned to one of three experimental conditions intended to vary perceived solidarity from women globally (high vs. low vs. control; see OSM). Participants then completed measures of betrayal and hawkish policy preferences regarding the war in Gaza. Because the manipulation did not reliably affect outcomes, experimental condition was retained as a covariate in all analyses.

In this context, hawkish policy preferences were assessed via support for forceful wartime measures and can be understood as an ingroup-defensive orientation; given that prominent women’s rights and feminist actors publicly foregrounded civilian protection and de-escalation during the war (International Women’s Development Agency [IWDA], 2023). Thus, stronger hawkish preferences can be interpreted as psychological distancing from the SOI with women worldwide.

Measures

Unless otherwise specified, responses were provided on 7-point Likert scales (1 = do not agree at all, 7 = agree to a large extent). Measures targeted women worldwide; parallel measures targeting “international feminists” were also collected and are reported in the Online Supplement.

SOI (T1)

SOI was assessed with respect to women worldwide with four items (e.g., “I feel close to women in other countries,” “I stand with women worldwide”; α = .88).

Expectations of Solidarity (T1)

Five items assessed expected support from women worldwide (e.g., “I expect women worldwide to take a stand against gender-based violence”; α = .93).

Intergroup Betrayal (T2)

Betrayal was assessed using the same five-item scale as in our previous studies, referring to women worldwide (α = .91).

Hawkish Policy Support (T2)

Support for hawkish wartime policies was assessed with five items (e.g., “I support increased military action in Gaza even if civilians are harmed”; α = .84).

Control Variables

We controlled for age, political ideology, religiosity, education, income, feminist ideology, and experimental condition.

Results

SOI (T1) served as the independent variable, expectations (T1) as the first-stage mediator, betrayal (T2) as the second-stage mediator, and hawkish policy preferences (T2) as the dependent variable.

Participants reported moderately high SOI with women worldwide (M = 5.10, SD = 1.25) and high expectations of solidarity (M = 6.23, SD = 1.12). At Time 2, which represents a time of threat to SOI, betrayal was also relatively high (M = 5.00, SD = 1.50), and support for aggressive war policies toward Gaza was around the scale midpoint (M = 3.39, SD = 1.53). All descriptive statistics and correlations across variables are presented in the OSM. As predicted, SOI (T1) was positively associated with expectations for solidarity (T1), r = .34, 95% CI [0.26, 0.41], and expectations were positive associated with betrayal under threat (T2), r = .30 [0.23, 0.38]. Moreover, there was a positive association between betrayal (T2) and hawkish policy preferences (T2), r = .23 [0.15, 0.31].

Sequential Mediation Model

We tested the hypothesized sequential mediation model using structural equation modeling (lavaan in R), estimating paths from SOI (T1) to expectations (T1) to betrayal (T2) to hawkish policy support (T2), controlling for age, education, income, religiosity, political ideology, feminist ideology, and experimental condition. The model showed adequate fit, χ²(1) = 5.03, p = .025, and explained 45% of the variance in hawkish policy support. Regression estimates are reported in Table 4.

Table 4.

Sequential Mediation: Dependent Variable Model for the Mediation of SOI Shared With Women via Expectations and Betrayal.

Predictor	B (SE)	95% CI	β	z	p
Age	−0.01 (0.00)	[−0.02, −0.01]	−.11	−3.19	.001**
Education	−0.09 (0.04)	[−0.17, −0.01]	−.08	−2.28	.023*
Income	0.02 (0.04)	[−0.05, 0.09]	.01	0.51	.611
Religiosity	0.16 (0.04)	[0.08, 0.24]	.15	4.15	<.001***
Political ideology	0.46 (0.05)	[0.36, 0.56]	.37	9.01	<.001***
Feminist ideology	−0.26 (0.05)	[−0.35, −0.16]	−.23	−5.24	<.001***
Condition: High Threat	−0.02 (0.12)	[−0.26, 0.21]	−.01	−0.16	.873
Condition: Control	−0.09 (0.11)	[−0.32, 0.13]	−.03	−0.81	.417
SOI T1	−0.08 (0.05)	[−0.17, 0.02]	−.06	−1.53	.127
Expectations T1	0.01 (0.06)	[−0.10, 0.14]	.01	0.21	.835
Betrayal T2	0.18 (0.04)	[0.10, 0.26]	.18	4.66	<.001***
R 2	.448
Observations	584

Note. 95% CI indicate percentile bootstrap confidence intervals (5,000 resamples) for the unstandardized coefficient. Reference category for condition is Low Threat. Estimation method: maximum likelihood with full information maximum likelihood for missing data. B = unstandardized regression coefficient; CI = confidence interval; β = standardized coefficient; SOI = superordinate identity.

*

p < .05. **p < .01. ***p < .001.

Consistent with predictions, SOI (T1) related to stronger expectations (T1), a₁ = 0.20, SE = 0.05, z = 4.31, p < .001, 95% CI [0.11, 0.29]. Expectations (T1) was associated with higher betrayal (T2), a₂ = 0.50, SE = 0.07, z = 7.66, p < .001 [0.38, 0.63]. Betrayal (T2) predicted stronger hawkish policy support (T2), b₂ = 0.18, SE = 0.04, z = 4.66, p < .001 [0.10, 0.26], whereas expectations did not show a direct association with hawkish policy support, b₁ = 0.01, SE = 0.06, z = 0.21, p = .835 [−0.10, 0.14].

The hypothesized sequential indirect effect from SOI to hawkish policy support via expectations and betrayal was significant (indirect effect = 0.02, SE = 0.01, z = 3.04, p = .002, 95% CI [0.01, 0.03]). The indirect effect from SOI through expectations alone showed no significance (indirect effect = 0.00, SE = 0.01, z = 0.20, p = .84 [−0.02, 0.03]), and nor did the total indirect effect (total indirect effect = 0.02, SE = 0.01, z = 1.63, p = .10 [0.00, 0.05]). Thus, SOI predicted greater hawkish policy support primarily indirectly through heightened expectations and subsequent betrayal. We also conducted a mediation sensitivity analysis for the indirect effect of solidarity expectations on hawkish policy support via betrayal; the indirect effect would be reduced to zero at ρ = .20 (OSM, Tables S27–S28).

Discussion

Study 3 extended the model to a cross-national SOI and a more institutionalized form of hardline political expression. Stronger superordinate identification at Time 1 predicted stronger expectations of solidarity from women worldwide, which in turn predicted greater betrayal in the Time 2 threat-salient context. Betrayal then predicted stronger support for hawkish wartime policies, controlling for political and feminist ideology, religiosity, and experimental condition.

Notably, the hawkishness outcome targeted a different outgroup (Palestinians in Gaza) rather than the perceived betraying group. This pattern aligns with work on spillover and displaced revenge (Chou et al., 2022; Stackhouse et al., 2023), suggesting that betrayal in one relational domain may increase willingness to endorse harm toward another target. More broadly, stronger hawkish preferences may reflect symbolic distancing from the superordinate community that was previously a source of identification, to the extent that participants perceived prominent international women’s rights and feminist actors as aligned with restraint-oriented positions during the war (IWDA, 2023). Taken together with Studies 1 and 2, Study 3 supports the broader claim that SOI threat can foster violent or hawkish responses via betrayal and clarifies one pathway through which this process emerges: heightened expectations at baseline that, when violated, give rise to betrayal.

General Discussion

In this work, we investigated a central issue in modern societies: What happens when minority groups who once felt included or recognized as part of a broader community come to perceive their inclusion as threatened? Moving beyond research on the positive impact of superordinate identities (SOIs), we examined their potential drawbacks under threat, conceptualizing SOI threats as exclusionary events that signal that minorities are not accepted by the majority with whom they shared a common identity. Across Studies 1 and 2, SOI threat was associated with stronger support for political violence among Ethiopian Jews and Arab-Muslims in Israel. As theorized, these links were strongest among those previously more invested in the superordinate “we,” and intergroup betrayal emerged as a key mediating variable. Study 3 extended this pattern to a transnational gender-based context, showing that identification with women worldwide was linked to later hawkish policy support indirectly via expectations of solidarity and subsequent betrayal under conditions that emphasized threat to SOI.

The variety of intergroup contexts suggests that the proposed process applies across settings. Study 1 examined an ethnic minority (Ethiopian Jews) with strong baseline connections to the majority via shared religious affiliation. Study 2 examined a national minority (Arab-Muslims) with more contentious relations with the majority. Study 3 extended the framework to a gender-based identity associated with solidarity expectations. Across studies, betrayal was assessed in parallel ways and linked perceived SOI threat to hardline attitudes. In Study 3, hardline responding was expressed through support for forceful wartime policies toward a different outgroup (Palestinians in Gaza)—a response that may also signal political distance from the global superordinate community. Relatedly, work on spillover and displaced revenge suggests that betrayal in one relational domain can heighten willingness to endorse harm toward related or symbolically connected targets (Bouman et al., 2014; Stackhouse et al., 2023). Thus, threatened SOI may foster support for hardline policies while also reflecting distancing from the broader “we.”

Notably, we were unable to produce reliable experimental variation in perceived SOI threat in the two studies that included a manipulation (Studies 1 and 3). Given these patterns, we tested the hypotheses using correlational data (retaining condition as a covariate). This difficulty may reflect that SOI threat is not easily induced via brief laboratory messages and may be more tightly linked to natural occurrences (Hopkins et al., 2020). Accordingly, assessing responses before and after potentially threatening events—as in Study 2—may offer a more promising approach.

Implications for Theory

Our research contributes to ongoing debates about the longer-term effects of SOIs on intergroup harmony (Cakal et al., 2011; Dixon et al., 2010; Saguy, 2018; Tropp & Pettigrew, 2005). Although unifying (Beaton et al., 2008; Dovidio et al., 2010), SOIs can foster relationship-based expectations of inclusion and solidarity that may not always be upheld. We advance the field by identifying intergroup betrayal as a mechanism linking SOI threat to hardline political attitudes and by showing that SOI threat can be especially consequential for those who have invested in the shared “we.” In doing so, we connect work on common ingroup identity and contact with research in political psychology and political science showing that discrimination and exclusion can mobilize minorities politically, sometimes in civic and sometimes in more hardline directions (Adam-Troian et al., 2020; Bilodeau et al., 2023; Oskooii, 2020; Weiss et al., 2023).

Extending these perspectives, our framework suggests that when exclusion is experienced as a violation of identity-based promises, betrayal may provide the affective pathway through which mobilization takes hardline forms. Given widespread use of common-identity approaches (Paluck et al., 2021), our findings suggest interventions should attend not only to cultivating a shared “we,” but also to the expectations such appeals create. This implies pairing common-identity messaging with credible and observable inclusion and follow-through from institutions, reducing the risk that solidarity promises are experienced as symbolic or broken (Gaertner & Dovidio, 2000; Saguy et al., 2009).

Beyond majority–minority relations, perceived exclusion can also shape how minority groups orient toward other stigmatized groups and coalitional partners. Work on stigma-based solidarity suggests that discrimination can sometimes foster coalitional alignment via a shared disadvantaged identity, but can also undermine solidarity and prompt defensive distancing when threat is construed as group-specific or competitive (Craig & Richeson, 2016; Liu & Carrington, 2022). Relatedly, political science research shows that exclusionary cues and cross-minority solidarity considerations can shape partisan alignments and intergroup coalitional preferences (Kuo et al., 2017). In this light, our findings suggest that betrayal, rooted in violated relationship-based expectations implied by a shared “we,” may be one affective pathway through which “broken solidarity” shifts responses from coalition-oriented inclusion to defensive hardening, including hardline attitudes that are not necessarily directed only at the original source of exclusion.

Because betrayal implies prior trust and relationship-based expectations (Finkel et al., 2002; Fitness, 2001; Lee & Selart, 2015), it differs from emotions often emphasized in work on radicalization, such as anger (Obaidi et al., 2020), contempt (Tausch et al., 2011), or humiliation (Webber & Kruglanski, 2018). Betrayal captures a relational rupture—an experience that “we had a bond, and you violated it”—that can convert exclusion into hardline political expression. In our studies, intergroup betrayal reflects a rupture in norms believed to govern a meaningful relationship built upon a shared SOI. Accordingly, strong prior commitment to an SOI may amplify readiness for hardline responding when exclusionary signals are interpreted as a breach of the relationship implied by that shared identity.

Our findings also speak to the integration paradox in political psychology, whereby more integrated individuals sometimes show stronger reactivity to discrimination and, in some cases, greater support for hardline responding (Lajevardi et al., 2020; Obaidi et al., 2019; Park et al., 2013). In both our longitudinal study with Arab-Muslims and our study with Israeli Jewish women, those who initially felt more included in a shared SOI were most likely to respond to exclusionary events with betrayal and hardline political expression. High SOI may thus create a psychological “height” from which to fall: when equality and solidarity fail to materialize, the gap between what was promised and what is delivered may be experienced as betrayal.

Finally, this insight underscores a key distinction between interpersonal and intergroup betrayal. In interpersonal contexts, strong commitment often fosters forgiveness (Finkel et al., 2002). Our findings suggest that in intergroup settings, high SOI and relational expectations do not prevent hardline responding following betrayal, and in fact might increase it (Rusbult & Van Lange, 2003). Betrayal can therefore reorient intergroup relations from cooperation toward conflict, including shifts in policy preferences that distance individuals from norms associated with the superordinate group.

Limitations and Future Directions

This research should be interpreted in light of several limitations. First, our models are observational and, in Study 1, cross-sectional; thus, we cannot make causal claims about the mediating role of betrayal. Although the over-time designs in Studies 2 and 3 strengthen inference by separating baseline predictors (SOI and expectations at T1) from betrayal and outcomes at T2, postthreat assessments were only made once. As Bullock et al. (2010) note, mediation analyses with observational data rest on strong assumptions, particularly that the mediator is not correlated with unmeasured causes of the outcome. We sought to mitigate this by adjusting for a broad set of covariates and exploiting temporal separation in Studies 2 and 3, but omitted-variable bias remains possible. Accordingly, the indirect effects should be interpreted as evidence consistent with the proposed mechanism rather than definitive causal estimates. Future work could leverage longer-term longitudinal designs around real-life threats and repeated postthreat assessments to test whether hardline reactions peak, fade, or escalate and which factors predict repair versus persistence (Finkel et al., 2002; Hannon et al., 2010).

Research on intergroup betrayal may also benefit from examining a broader range of political outcomes beyond support for violence or hawkish policy preferences, including constructive collective action (Tausch et al., 2011; Van Zomeren et al., 2008) and electoral participation (Besco, 2024; Weiss et al., 2023). Such outcomes may help distinguish when betrayal fuels withdrawal versus mobilization, and when responding becomes civic versus hardline. Future work might also examine betrayal among higher-power groups, for whom greater capacity for retaliation may translate betrayal perceptions into punitive policies or escalatory conflict behavior. Finally, future research could more directly compare betrayal with other intergroup emotions (e.g., anger, humiliation) and test whether betrayal uniquely predicts hardline outcomes beyond these related affective responses (Mackie et al., 2016).

A further limitation concerns our samples and research settings. Study 1 relied on snowball sampling of Ethiopian Jews, and both Studies 1 and 2 drew on relatively small, nonprobabilistic samples in a single national context. Study 2 and Study 3 were also conducted around highly charged political events, and all studies suffered from some degree of attrition. These features limit the generalizability, precision, and statistical power of our estimates, particularly for smaller effects, which can make some associations harder to detect reliably. Accordingly, these studies are best suited to testing theoretically specified relationships and mechanisms in ecologically valid contexts, rather than estimating population parameters with high precision. Because recruitment was nonprobabilistic, unobserved selection processes may have shaped who entered the sample, which could have affected correlational estimates. At the same time, Ethiopian Jews, Arab-Muslims in Israel, and Israeli Jewish women in the immediate aftermath of October 7 represent hard-to-reach populations and moments that are rarely accessible for systematic survey research (Leshem et al., 2020). Our data therefore offer a rare window into how real-world exclusionary turns and perceived failures of solidarity are experienced by minority communities that had invested in shared identities with majority or transnational partners. Future work using larger, probability-based samples and comparative designs could test the robustness of these patterns across institutional contexts and minority groups, while still capitalizing on the ecological validity that such hard-to-reach settings provide.

Conclusion

Across studies of Ethiopian Jews, Arab-Muslims, and Israeli Jewish women, intergroup betrayal—stemming from unmet expectations and ruptured trust within superordinate identities—emerged as a key mechanism linking perceived SOI threats to support for violent intentions and hawkish policy preferences. These findings underscore that shared identities can be both a resource and a vulnerability when their implied promises are perceived as broken.