Postconflict reconciliation efforts around the world have stumbled upon persistent division among former antagonists, long after formal violent confrontation has ended (Gibson Reference Gibson2004; Maddison and Shepherd Reference Maddison and Shepherd2014; Tam et al. Reference Tam, Hewstone, Kenworthy, Cairns, Marinetti, Geddes and Parkinson2008; Hayward Reference Hayward, Ugarriza and Caluwaerts2014; Halperin Reference Halperin2011).
Negative intergroup attitudes, discourses, and emotions block the reestablishment of harmonious relations between former antagonists (Rettberg and Ugarriza Reference Rettberg and Ugarriza2016). One major strand of literature explains postconflict intergroup hostility as a consequence of the failure of accountability mechanisms to deal with past trauma, victimization, and perceived harm. Bridging divides would require a broad societal process of coming to terms with the immediate past of violence (Encarnación Reference Encarnación2008), sometimes with the implementation of formal mechanisms of justice, truth, reparation, or memory (Aiken Reference Aiken2008; Clark & Kaufman, Reference Clark and Kaufman2009; Gellman Reference Gellman2008; Ure Reference Ure2008).
Another strand focuses less on effects from past experiences and more on ongoing sources of prejudice and subsequent biases. Postwar hostility here would be a consequence of persistent war mentalities that typically dehumanize the other, perceiving them as undesirable or even as a threat (Abu-Nimer Reference Abu-Nimer2001). Improvement of intergroup attitudes would require a transformation of individuals’ dispositions and perceptions, which may or may not require dealing with past trauma (Brounéus Reference Brounéus2010; Duncan Reference Duncan2009; Hirsch Reference Hirsch2012).
This article addresses the question of whether unaddressed war grievances, such as victimization and or legacies of ideological radicalization, better explain the divide between ex-military and former guerrillas in Colombia, as expressed in their intergroup bias. As such, it aims to explore the extent to which ideological attitudes along the left-right spectrum explain the observed implicit bias. Understanding the structure underlying postconflict antagonisms should let us fine-tune public policy and societal efforts aimed at overcoming obstacles to reconciliation.
To answer such a question, we sampled two pools of ex-combatants who actively participated in the armed conflict between 1996 and 2016: one of 697 Colombian Army soldiers in the process of retirement and another of 394 former guerrillas in the process of reintegrating after demobilization. A total of 484 voluntary participants took part in the research.
Sources of postconflict divide
Some of the most robust contributions attempting to discern the specific contours and sources of the observed intergroup hostility in divisive contexts come from studies on the effect of victimization (Sinayobye et al. Reference Sinayobye, Hameiri, Vollhardt, Nadler and Vollhardt2020) or “perceived harm” (Bilali et al. Reference Bilali, Tropp and Dasgupta2012). According to this literature, it is the subjectively perceived grievances of some parties that negatively condition attitudes and emotions. The fact that all wartime parties tend to claim victim status and blame their antagonists as perpetrators only complicates matters (Shnabel et al. Reference Shnabel, Halabi and Noor2013; SimanTov-Nachlieli et al. Reference SimanTov-Nachlieli and Hakabim2015). Under this paradigm, resolving the problem of victimization status lies at the root of any perspective of postconflict reconciliation.
Empirical works add weight to this approach, providing evidence that wartime exposure to traumatic experiences helps explain observed postconflict intergroup hostility in African cases (Bayer et al. Reference Bayer, Klasen and Adam2007). In Colombia, non-ex-combatant war victims tend to have a negative bias toward ex-combatants (Ugarriza et al. Reference Ugarriza, Natalia Trujillo-Orrego, Ortiz-Ayala, Rodriguez-Calvache and Quishpe2022). A proposed mechanism for explaining such outcomes describes how higher levels of war-related trauma are associated with distinctive behavior and biased information processing (Gómez et al. Reference Gómez, López Hincapié, Cardona, Ugarriza, Herrera and Trujillo2022; Trujillo et al. Reference Trujillo, Giraldo, López, Acosta and Trujillo2021), as well as mental health outcomes including anxiety disorders, risk of suicide, and post-traumatic stress disorder (Trujillo et al. Reference Trujillo, Trujillo, Valencia, Ugarriza and Acosta Mesas2019). In the same vein, soldiers from the Colombian Army who considered themselves to be victims tended to show higher levels of post-traumatic stress disorder, hypervigilance, aggression, and emotional reactivity, as well as lower levels of empathy, when compared to nonvictim veterans (Gantiva et al. Reference Gantiva, Suárez-Pico, Aristizabal-Gómez, Granada-Aguirre, Suárez-Lara, Tenorio-Quiñones, Arias-Higuera, Guzmán-Durán, Castiblanco-Durán and Hurtado-Parrado2023; Ugarriza et al. Reference Ugarriza, Trujillo, Trujillo and Acuña2025).
In contrast with these findings, Kao and Revkin (Reference Kao and Revkin2023) argue that perceptions of the antagonists’ motivations and behaviors might actually better explain intergroup hostility than actual harm or victimization experiences. Even in the absence of subjective victimization, individual attitudes become embedded in divisive narratives that reinforce prejudice and bias against the antagonists, who are perceived as threats even when violence has decreased or stopped altogether (Radnitz Reference Radnitz2018; Petrović et al. Reference Petrović, Mededović, Radović and Radetić2019).
Beyond psychological and physical harm for veterans (Weiss et al. Reference Weiss, Fletcher and Hendricks Thomas2023), a major legacy of politically motivated violence is the reshaping of identities and the deepening of ideological radicalization that may even be transmitted across generations (Lupu and Peisakhin Reference Lupu and Peisakhin2017). As such, the persistence of ideological conflict (e.g., religious, nationalist) correlates with a weaker disposition to reach a compromise with the antagonist, even beyond material considerations (Canetti et al. Reference Canetti, Rubin, Khatib and Wayne2019). In Colombia, authors have even found that political and ideological preferences, rather than exposure to conflict-related experiences, better explain people’s attitudes toward peace-related issues (Liendo and Braithwaite Reference Liendo and Braithwaite2018).
Ideology and radicalization
Ideology can be understood as a set of beliefs and values that guide individuals’ perceptions and attitudes toward the world around them (Gerring Reference Gerring1997). This broader concept can be framed as a corpus of thought that organizes specific elements commonly present in armed conflicts, such as doctrines, narratives, symbols, and myths (Gutiérrez and Wood Reference Gutiérrez and Wood2014). Gerring explains that scholars often add context-specific attributes to clarify their use of the term. These attributes help specify whether the concept is being used in a broad or narrow sense, and in some cases, ideology may refer to thought, language, behavior, or all these phenomena at once.
Modern approaches to the role of ideology in contemporary conflicts suggest that while ideology is not a primary motivation for joining armed groups, it is necessary to incorporate a comprehensive “need, greed, and creed” framework to better understand the emergence of violent conflict (Zartman Reference Zartman, Cynthia and William Zartman2005).
In the case of Colombia, studies have demonstrated that warring factions are clearly divided along ideological lines (Ugarriza and Craig Reference Ugarriza and Craig2013), and ideology has significantly influenced the dynamics and motivations of the military and the guerrillas (Ugarriza and Pabón Reference Ugarriza and Pabón2017).
Much empirical research on ideology has employed discourse as the main unit of analysis. Many studies on ideology and discourse are qualitative in nature (Foucault [Reference Foucault, Burchell, Gordon and Miller1984] 1991; Howarth Reference Howarth, Howarth and Torfing2005; Van Dijk Reference Van Dijk1993), although some quantitative approaches use statistical, logical, or lexicographic methods (Myhill Reference Myhill2005).
In addition to discourse, scholars have also examined the strategic use of symbols and other cultural references to stir emotions and mobilize groups for violence and military action. Combatants’ responses to these forms of communication are partly shaped by emotional reactions. Hostile feelings toward enemy groups can indeed hinder rational behavior and promote violence (Figueiredo and Weingast Reference Figueiredo, Weingast, Barbara and Snyder1999; Kaufman Reference Kaufman2006; Mertus Reference Mertus1999; Petersen Reference Petersen2002).
Another dimension of ideological commitment is reflected in explicit attitudes. In this manuscript, we focus specifically on attitude measures to assess the ideological positions of leftist ex-guerrillas and right-wing military personnel. Using a set of statements designed to evaluate attitudes on topics like violence, the economy, and public policy, we gain a clear picture of how individuals perceive and respond to their ideological contexts.
The decision to concentrate on attitudes rather than extending the analysis to discourse and emotions is grounded in prior research (Ugarriza and Craig Reference Ugarriza and Craig2013), which has shown that it is possible to capture the essence of ideology without measuring all its dimensions simultaneously. Discourse, emotions, and attitudes, therefore, represent different dimensions of a single latent variable.
The attitude scales consist of five items that capture favorable or unfavorable opinions toward rival ideological groups. Each item assesses agreement or disagreement with statements that assert hypothetical positive or negative effects of leftists or rightists on issues such as violence, economic performance, democratic participation, political debate, public expenditure, and the overall strength of society. The statements are designed to determine whether different groups can be distinguished on the basis of their rationalized positions toward perceived adversaries.
These attitude measures, together with victimization reports, serve as alternative explanatory factors for our dependent variable, namely the intergroup bias measure.
Colombia as a relevant case
In Colombia, the series of peace agreements signed between armed groups and governments from 1989 to the present has not generally translated into harmonious relationships between former combatants, victims, and host communities (Prieto Reference Prieto2012). Furthermore, antagonisms related to war have translated into everyday prejudice—as observed with implicit and explicit biases—and mistrust in postconflict contexts, manifesting in turn in social and ideological gaps and polarization (Nussio et al. Reference Nussio, Rettberg and Ugarriza2015).
Although they were not the only actors at play, the Colombian armed conflict was primarily fought between state armed forces and Marxist guerrillas engaged in fighting, both of which historically understood the conflict not only in social and economic terms, but also as an ideological confrontation.
For decades, the Colombian Armed Forces’ military doctrine described guerrilla warfare as just one modality of many by which insurgents attempted to “impose puppet governments” at the service of “totalitarian powers’ interests” (Colombian Army Command 1983, 6). Insurgent groups were systematically described as the “armed wing” of the insurgent forces, which also counted on the “insurgent civil population” as the nonmilitary wing (Colombian Army Command 1987, 20).
Military textbooks included “education for democracy” courses for soldiers as part of their basic training, aiming to “demonstrate and convince soldiers that solutions to the country’s problems [are] not Marxism but democracy” (Colombian Armed Forces 1988, 13). The idea of fighting a “guerrilla revolutionary warfare” persisted in military manuals way beyond the twenty-first century (Colombian Armed Forces 2010). This ideological understanding of the armed conflict was prolonged after the biggest Marxist guerrilla group signed a peace accord in 2016 and demobilized, and the military continued seeing its former enemies as a political adversary (Castillo and Niño Reference Castillo and Niño2020).Footnote 1
In the case of the former guerrillas who demobilized in 2016, the peace accord meant their transformation into a legal political party and participation in the legal political system, without abandoning their wartime hierarchies, agendas, and ideological stances. Their communist orientation was well defined in established goals, such as to “overcome the capitalist social order [now] in force in Colombian society … [and] the construction of a new political economy” (FARC 2016).
In the next sections, we systematically test whether ideological stances help explain the persistence of the military-guerrilla divide, and we compare their explanatory power with that of self-perceived victimization experiences.
Methods
Participants
We conducted a two-stage large-scale fieldwork effort to create samples of professional soldiers in the process of retirement and ex-guerrillas who had just disarmed and were enrolled in the early stages of the official reintegration program. The process of decommissioning military and guerrillas was a product of the peace agreement between the Colombian government and the FARC guerrillas in 2016, and both groups were subjected to differentiated programs of reintegration into civil life.
Between 2016 and 2017, the Colombian Armed Forces implemented a program of assisted retirement for approximately six thousand soldiers as part of a troop reduction plan. This program took place in military bases throughout the country, where soldiers gathered and received educational, health, and labor-related services in preparation for their transition back to civilian life. In that period, the program assisted around three thousand soldiers per year. In 2017, there were a total of 3,426 attendees, 98 percent of whom came from the army and the rest from the navy and air force. Such figures resemble those of the actual active military population in 2017, where about 83 percent of soldiers were enlisted in the army, 12 percent in the navy, and 5 percent in the air force.
As part of this larger program, we visited sixteen municipalities and created a pool of soldiers in the process of retirement, applying a clustering and randomization procedure. The first step consisted of identifying the eight major theaters of operation (i.e., divisions) in which the army has historically divided the country. By making sure we covered every cluster, we could reasonably gather a proper representation of soldiers who were present in different geographical zones at the time of the survey. As a second step, in each of the clusters, we randomly chose to visit at least one urban and one rural-based military installation (two bases on average per Division). In every location, the army arranged a one-day session with all soldiers in the process of retirement so they could participate in our survey. As a result, we were able to visit sixteen installations from a total of twenty-four where the program was being held, and we created a pool of 697 soldiers in the process of retirement, with national representativeness.
Participants in our pool, all rank-and-file professional soldiers, were on average thirty-nine years old with twenty years of military service and eight years of formal education; no women were included (female professional soldiers are rare to nonexistent in the Colombian Armed Forces), and 69 percent of the sample declared to have been war victims. Professional soldiers were typically previously drafted fighters who decided to continue a paid military career and were specifically trained to be part of the frontline forces against guerrillas. Among all strata of the armed forces, they represented the core of the actual battlefield combatants. This is to say, our sample does not represent the army as a whole, but it does represent those soldiers who were offensively deployed on the battlefield between 1993 and 2016.
One potential bias that we had foreseen was that soldiers’ deployment zones did not necessarily match those where they joined the retirement program. To control for this variation, we collected individual-level data on the main military unit they made part of during wartime, which let us establish with precision the zone in which they actually fought.
In 2017, as the second stage of our fieldwork, we started visiting the demobilization zones of former FARC guerrillas, where they began their reintegration process to civilian life, after completing an UN-supervised disarmament process that same year. Officially, there were twenty-six such demobilization zones in 2017 all over the country, although some ex-combatants opted to move to nearby towns and cities right after demobilization.
Similar to the military sampling, we made sure to cover all major regions of the country so that we could create a pool reflecting the cultural and geographical variance of the former guerrillas. Also, since most guerrillas demobilized in zones near their zone of military operations, we also reasonably accounted for membership in all major FARC structures.Footnote 2
A small group of former mid-level FARC members was tasked with contacting leaders in each encampment zone to extend an invitation to participate in the study. After the invitation, our research team personally visited each zone and worked with voluntary participants. No financial compensation was offered. However, the FARC political party was permitted to later make use of the collected data for their own internal analysis.
A total of 394 ex-guerrillas accepted to be part of our pool. Approximately 30 percent of them had fought in the guerrilla group’s eastern bloc; 12 percent in the southern bloc; 7 percent from Magdalena Medio; 6 percent in the western bloc; 3 percent in the northwestern bloc; three more in the central bloc; and 14 percent in the northern bloc. Approximately 24 percent of ex-combatants did not disclose their former military structure, although their reintegration zone might have served as an indication. All these ex-combatants were found along the five major geographical zones described earlier.
Participants in our pool of ex-guerrillas had an average age of forty years old, seventeen years of wartime experience, and ten years of formal education (close to high school graduate level). Forty-two percent were women, and 70 percent declared themselves to have been war victims.
We are aware of potential sources of nonrepresentativeness in our pools. Soldiers and guerrillas who died in the course of the conflict, or those who deserted, are clearly not included in our study. We can only guess, at best, that combatants and veterans represented in our sample were exposed more or less to the same kind of risks as those nonrepresented. Inevitably, interpretations of our results need to be assessed with these limitations in mind.
All 1,091 members of our pools were invited to participate in the following stage of our research. A total of 484 accepted our invitation to spend about thirty minutes filling out our questionnaires and completing the computer-based tests. We include a statistical description of the sample in Table A1 of the supplementary material. We found a significant correlation between more years spent at war and fewer years of education, on the one hand, and a higher propensity to accept our invitation, on the other. These two variables will be controlled for in the following analyses.
Measures and procedure
We wanted to analyze the extent to which ideology could explain intergroup biases between former war enemies in postconflict settings, beyond grievances derived from previous victimization. Focusing on a continuous measure of bias, rather than binary group membership, we can capture not only the capacity of our independent variables to tell the differences between the two groups but also the intensity of their divide. Thus, our dependent variable is defined as the measure of bias between former soldiers and guerrillas.
Instead of asking directly in our surveys, we relied on a computer-based task known as the Implicit Association Test (IAT) to conduct such a measure.
Implicit attitudes, or implicit biases, are prejudice-led, memory-based, and effortless neural resources that produce automatic responses and are beyond conscious control (Banaji and Greenwald Reference Banaji and Greenwald1995; Banaji et al. Reference Banaji, Lemm, Carpenter, Tesser and Schwartz2001; Bargh Reference Bargh, Robert and Wyer1997; Greenwald et al. Reference Greenwald, McGhee and Schwartz1998; Neumann et al. Reference Neumann, Hülsenbeck and Seibt2004; Pérez Reference Pérez2010; Wilson et al. Reference Wilson, Lindsey and Schooler2000). As a result, they capture a different aspect of bias compared to explicit measures, which reflect consciously controlled beliefs (Fazio and Towles-Schwen Reference Fazio, Towles-Schwen, Chaiken and Trope1999; Fazio and Olson Reference Fazio and Olson2003). In fact, it is not uncommon for implicit attitudes to contradict an individual’s self-reported view of the same target (Dovidio et al. Reference Dovidio, Kawakami, Beach, Brown and Samuel2001; Devos and Banaji Reference Devos and Banaji2005; Arcuri et al. Reference Arcuri, Castelli, Galdi, Zogmaister and Amadori2008).
IATs elicit rapid responses from individuals by exposure to stimuli related to specific groups, individuals, or preferences (Greenwald et al. Reference Greenwald, Brendl, Cai, Cvencek, Dovidio, Friese, Hahn, Hehman, Hofmann and Wiers2021, Reference Greenwald, Nosek and Banaji2003). Previous studies have demonstrated the potential of IATs in understanding diverse social biases among ex-combatants and communities (Unfried et al. Reference Unfried, Ibañez and Restrepo-Plaza2022). Here, we use contrasting group categories “military” and “guerrilla.” Participants are told to associate each of these groups with images on a screen and audio-recorded valence words (e.g., good, bad). Participants’ response times serve as an indication of the presence or absence of biases, specifically when asked to associate group categories with positive or negative traits. A negative IAT score indicates a bias against guerrillas when compared to soldiers, and vice versa. The higher the score, the more positive the bias.Footnote 3 A visual description of the IAT task is provided in Figures A1 and A2 in the supplementary material. Also, we provide a detailed description of the process of selection and validation of valence-loaded terms in Table A2 in the supplementary materials.Footnote 4
When measuring ideology, individuals’ personal factors—such as education, income, and political experience—interact with broader political and cultural contexts to shape how left-right labels are interpreted (Zechmeister and Corral Reference Zechmeister and Corral2013). This interaction highlights the fluidity and complexity inherent in people’s understanding of these terms (Zechmeister Reference Zechmeister2006).
Several well-established strategies for measuring group-based ideological attitudes do not rely on individuals’ specific interpretations of ideological labels. Instead, they focus on social identity, emotional attachments, affective political polarization, and partisan stereotypes.
One common method is the use of feeling thermometers, with which respondents rate various political groups (e.g., “liberals,” “conservatives,” “leftist groups,” “rightist groups”) on a scale from 0 to 100, with 0 indicating very unfavorable feelings and 100 indicating very favorable ones.Footnote 5 Some surveys adapt this measure to assess whether respondents perceive these groups as beneficial or harmful to the country or to specific areas such as the economy.
Another approach involves Partisan Social Identity Scales, which measure in-group favoritism and out-group hostility toward ideological groups. Sample items include statements such as “Right-wing politicians care more about businesses than the general public” or “Leftist groups are good for economic equality but bad for economic growth” (Mason Reference Mason2018; Huddy et al. Reference Huddy, Mason and Aarøe2015; Greene Reference Greene1999).
Similarly, the Stereotype Content Model (SCM) (Fiske et al. Reference Fiske, Cuddy, Glick and Xu2002; Cuddy et al. Reference Cuddy, Fiske and Glick2008) assesses the warmth and competence stereotypes that people hold toward various social groups, including political ones. The two traits often reflect perceived status and competition between groups. For example, respondents may be asked whether “leftist groups are competent at managing the economy” or whether “right-wing groups lack warmth and care for people.”
Drawing on these approaches and modeling our instruments after the SCM framework, we developed our own batteries for measuring intergroup attitudes.
Our main independent variable captures ideological attitudes. A total of six interspersed Likert-type items were used for measuring attitudes toward leftist or rightist groups.Footnote 6 In each case, half the sentences were worded in positive terms and the other half in a more negative tone. Sentences aimed to capture biases in terms of who is to blame for political, economic, social, and security problems (leftists and rightists) and who is contributing to solving them. The battery asks “How much do I agree or disagree with the following statements?” and includes sentences such as “Leftist groups are good for the economy of the country,” and similar for rightists. See more details in Table A3 of the supplementary material. Items were aggregated to estimate scores on a scale from –6 to +6. We reversed coding negative statements to reflect “agree” or “agree strongly” as a negative stance when appropriate. Thus, more positive aggregated scores reflect more positive attitudes, and vice versa. Survey items are transcribed in Table A3, in the supplementary material.
As a complementary measure, we applied a commonly used ideology scale, where participants place themselves on a scale from 1 to 10: The lower the score, the closer one’s affinity to the political left, and vice versa.Footnote 7
Participants were given a questionnaire of basic demographic information (e.g., age, gender, education level, military unit); also, some items explored whether they considered themselves conflict victims and, if so, of whom. There, we included our explicit attitude instruments.
Right after finishing the written survey, participants were invited to take our Implicit Association Test. For that purpose, participants were placed in a separate room. Researchers then offered basic verbal instructions for what the test was about and helped them put on earphones for the audio-recorded stimuli.
Hypotheses
We expected that ideological stances should reflect on the implicit social bias. Thus, our set of hypotheses is as follows:
H1: Participants with positive attitudes toward the ideological left should tend to have a more positive bias against guerrillas.
H2: Participants with positive ideological attitudes toward the right should tend to have a more negative bias against guerrillas.
Our alternative set of hypotheses suggests that victimization should better explain the persistence of the observed military-guerrilla divide than ideological measures. In previous studies, victims have been identified on the basis of their own account of war-related traumatic experiences (Giraldo et al. Reference Giraldo, Camilo Aguirre-Acevedo, Trujillo, Ugarriza and Trujillo2020), and some are officially sanctioned as such by state support institutions (Gantiva et al. Reference Gantiva, Suárez-Pico, Aristizabal-Gómez, Granada-Aguirre, Suárez-Lara, Tenorio-Quiñones, Arias-Higuera, Guzmán-Durán, Castiblanco-Durán and Hurtado-Parrado2023). Here, we relied on self-reported victimization perpetrated by guerrilla groups, paramilitary, or state forces. We asked “Do you consider yourself or your family to be victims of guerrilla groups, paramilitary, or state forces? Which groups?” (See original items in Table A3 of the supplementary material).
Our alternative set of hypotheses is the following:
H3: Participants who consider themselves victims of guerrillas should tend to have a more negative bias against guerrillas.
H4: Participants who consider themselves victims of state forces should tend to have a more positive bias against guerrillas.
Results
As expected, Table 1 shows that intergroup biases between former soldiers and guerrillas are strong. We can see how the full sample tended to have longer response times, on average, when exposed to positive words toward guerrillas than to negative ones, and therefore, we report a positive difference (102 milliseconds). In the first column from the left, we see how, in the case of positive words toward the military, reaction times are slower on average than in the case of negative words, and we report a negative difference (–84 ms).
Military vs. guerrilla IAT scores and reaction times per group
Note: Since our test was administered to a low-education sample, we opted not to eliminate subjects for whom more than 10 percent of trials had a latency below 300ms, and added latencies of incorrect and correct trials for the estimation of individual IAT scores.
* Standard deviation in parentheses.
** Reaction times are estimated as an average of practice and test blocks. For ex-combatants/positive, and victim/negative, the average is calculated from the corresponding B3 and B4 trials. For guerrilla/negative and military/positive, the average is calculated from the corresponding B6 and B7 trials.
Average differences in reaction times are used to estimate IAT scores for each participant in the sample. For the full sample of subjects (484 observations),Footnote 8 we report an average IAT effect of –0.426. Signs indicate average implicit attitudes toward guerrillas (baseline), which are negative in the case of ex-soldiers (–0.650), and positive for ex-guerrillas (0.229). In the first column, we present the different comparisons conducted between subgroups in our sample. Differences in scores between factions are statistically significant (F = 398.28, ρ = 0.000; Cohen’s d = 2.101, CI: –2.346807, –1.854773). This is also the case between males and females (F = 87.80, ρ = 0.000; Cohen’s d = –1.401, CI: –1.707, –1.094), although no female soldier, only female ex-guerrillas, was actually included in the sample. Despite the fact that we did not find strong differences between self-reported victims and nonvictims in our sample (F = 4.32, ρ = 0.038; Cohen’s d = 0.199, CI: 0.108, 0.387), we did when comparing those reporting to have been victims of guerrillas and those who did not (F = 136.37, ρ = 0.000; Cohen’s d = –1.305, CI: –1.545, –1.063). Differences are also significant between those who considered themselves victims of state forces and those who did not (F = 31.49, ρ = 0.000; Cohen’s d = 0.695, CI: 0.446, 0.943). We found no significant correlations with IAT scores in the case of age, years of education, or wartime years.
Crucially, we wanted to analyze to what extent ideological attitudes and victimization helped to explain the observed intergroup social biases, as shown in the ordinary least squares (OLS) analyses. In the second column of Table 2, model 1 shows the extent to which belonging to one group or the other (i.e., military or guerrilla) explains the IAT measure of intergroup bias, controlling for age, gender, year of education, and wartime years. Although this first model does not test our main hypothesis, it is an important reference, as it confirms that our social bias measure also reflects group membership.
Ideological measures and Intergroup social bias (dependent variable)
Note: Standard errors in parentheses. Control variables in all models: age, gender, year of education, and wartime years. Breusch-Pagan/Cook-Weisberg test (for heteroskedasticity) χ 2 = 0.82, p = 0.366. Mean variance inflation factor (VIF): 2.13.
*p < 0.05. **p < 0.01. ***p < 0.001.
The second set of regressions (models 2–4) tests our main hypotheses by exploring the explanatory power of ideological attitudes toward the left (model 2), toward the right (model 3), and self-reported scores along a left-right continuum (model 4). Since the variables explain similar portions of the observed variance, they produce a collinearity effect when included together in a saturated model. Therefore, we present each variable in an independent model. A similar procedure is applied to the victimization measures. The following set of regressions (models 5–6) shows the relation between reporting to have been a victim of guerrillas (model 5) or state forces (model 6). Since we mainly want to compare the predictive power of ideology to that of reported victimization experiences, a nonsaturated model 7 incorporates both ideological and victimization measures.Footnote 9
Our first model accounts for the effects of the full experience of being part of one of the warring groups, past and present. As expected, we see how being a soldier is correlated with a negative opinion of ex-guerrillas, and vice versa.
In models 2–4, the ideology indicators have significant coefficients and point to the fact that, the better the attitudes toward leftists, the higher, and therefore more positive, the social bias score, and vice versa. We also find the reverse trend in the case of attitudes toward rightists. Substantially, the social bias against guerrillas is worsened on average by 0.067 score points for every point decrease in the scale of attitudes toward the leftists (model 2), and it also worsens on average by 0.054 for every point increase in the scale of attitudes toward the rightists (model 3). The social bias score also decreased by 0.060 points on average for every point increase in the ideological scale in the left-to-right scale (model 4). As we can see, these three indicators behave in a very similar fashion and could be used interchangeably.
In the case of victimization experiences, participants who declared to have been victims of guerrillas tended to worsen their negative bias against these groups by 0.619 score points on average compared to those self-declared nonvictims (model 5). Conversely, those who claimed to have been victimized by state forces tended to improve on average 0.291 points on the social bias measure when compared to nonvictimized participants (model 6). In consonance with previous literature, victimization once more helps to explain postconflict intergroup bias.
On model 7, we test whether ideological scores and victimization measures can potentially explain together different portions of the variance of the intergroup bias measure. There, we observe that a composite model, which includes both ideology and victimization measures, as well as an interaction effect, explains a significant amount of variance. Here, we can see how ideology and perceived victimization reinforce each other while still accounting for different portions of the observed social bias variance. Individuals with both an ideological attitude bias and a self-reported victimization status exhibit a significantly higher social bias.
Figure 1 illustrates the interaction effect between ideological attitudes toward leftist groups and victimization by guerrillas on intergroup social bias. The plot shows the predicted levels of intergroup bias across the ideological spectrum (ranging from 0 to 10), with separate lines representing individuals who were victims of guerrilla violence and those who were not. In both groups, ideological attitudes exhibit a proportional association with bias scores. However, among victims, the regression line flattens and remains close to zero, suggesting that the influence of ideology on intergroup bias is moderated by the experience of victimization.
Interaction effect of attitudes toward left and victim of guerillas
The positive slope of both lines indicates that ideological attitudes play a significant role in shaping intergroup bias, regardless of victimization.
Discussion
Ideological commitment and reported victimization experiences, as we have seen, account for most of the observed postconflict intergroup biases, while an interaction effect may signal that ideology and victimization feed each other.
Wartime ideology becomes a problem in postconflict settings. Attitudes, discourses, and emotions seem to stand in the way of reconciliation between factions. The nature of the divide between parties has to do not only with grievances and war experiences but also the effort of their leadership to create a pervasive divide, useful in wartime, but problematic in peacetime. Even after armed confrontation ceases, ideological warfare and adversarial mentality persist. Concluding that the social divide between ex-military and former guerrillas has an ideological content may not be an intuitive result for those who perceive contemporary armed conflicts, such as the one in Colombia, as lacking significant ideological motivations.
We have resorted to a rather uncommon measurement tool for social science research, namely the IAT. While the IAT is a valuable tool for uncovering implicit biases, we can see how explicit attitudes might be regarded as a more relevant approach. In future research, we suggest employing a mixed-methods design that includes implicit and explicit intergroup bias measures as dependent variables.
The external validity of the results depends on several factors. This study focuses on Colombia and ex-combatants, which may somewhat limit generalizability to other postconflict contexts or societies without a similar history of armed conflict.
However, while Colombia’s conflict has unique features, many dynamics, such as ideological radicalization and the effects of victimization, are common in other conflicts. These parallels may allow for partial generalization of the results to other postconflict scenarios, particularly in polarized societies marked by extreme ideologies and histories of violence.
One key question is whether this war-related divide also extends to noncombatant civilians, and can potentially be explained also in terms of victimization and ideology. In countries that suffered major armed conflicts, as in Colombia, we would expect a large proportion of citizens to have suffered its effects, including those on dispositions and attitudes here described. However, we would also expect future studies to analyze different effects of direct versus indirect exposure to conflict-related traumatic experiences and their relation to different levels of ideological attitudes.
While the results of the study presented here offer some microfoundational elements to understand the role of ideology in contemporary armed conflict and postconflict settings, they also have important consequences for policy interventions and peace-building initiatives.
The results challenge the notions of intergroup attitude improvement as an exclusive product of dealing with the immediate past of violence, that is, taking a vertical approach to the problem. While they do not undermine the importance of truth telling, justice, and victim reparation, they nevertheless support the relevance of looking at the postconflict intergroup hostility problem from a different angle.
Our results support policy approaches that stress the need for attitudinal or emotional change as a condition to reestablishing relations between former antagonists (Bar-Tal and Bennink, Reference Bar-Tal, Bennink and Bar-Siman-Tov2004; Brounéus Reference Brounéus, Ambos, Large and Wierde2009; Bruneau and Saxe Reference Bruneau and Saxe2012; Long and Brecke Reference Long and Brecke2003; Nadler et al. Reference Nadler, Malloy and Fisher2008; Poitras Reference Poitras2010). In practical terms, an important strand of literature, particularly that drawing from the contact hypothesis (Pettigrew et al. Reference Pettigrew, Tropp, Wagner and Chirst2011), has pointed out ways that attitudes and dispositions could be changed for the better in postconflict societies all over the world (Tropp et al. Reference Tropp, Hawi, O’Brien, Gheorghiu, Zetes and Butz2017; Wright et al. Reference Wright, Tropp and Mazziotta2017).
Our findings imply that interventions targeting ideology—such as political education, intergroup dialogue, or exposure to alternative perspectives—may still be effective in reducing bias, even among those who have experienced victimization.
Although victimization appears to make intergroup bias more persistent, the results suggest that ideological interventions remain a promising strategy for fostering reconciliation. Therefore, peace-building efforts should not overlook the role of ideology. Instead, they should focus on shaping ideological attitudes to promote intergroup understanding. Programs emphasizing shared values, common economic and social goals, and historical reconciliation could help victims reframe their biases. Additionally, media, education, and political discourse play a critical role in reinforcing or challenging divisive narratives, making them essential components of any comprehensive reconciliation strategy.
We provide fresh evidence to support the normative claims around the relevance of ideology to contemporary politics. The conclusions regarding how ideology and polarization persist after conflict could also apply to other countries with enduring ideological tensions, even outside the context of war. This should contribute to a broader understanding of how political and ideological identities shape reconciliation and intergroup relations in civil society.
While this paper focuses on the roles of ideological attitudes and victimization experiences in shaping intergroup biases, other factors also contribute to divisions between factions. Intergroup biases emerge from a complex interplay of social, political, and cognitive elements, which can amplify divisions and fuel negative attitudes.
Intergroup contact, or the lack thereof, plays a pivotal role. Positive interactions between groups reduce prejudice, whereas negative encounters or the absence of contact deepen biases. Pettigrew and Tropp’s (2006) meta-analysis demonstrates that intergroup contact fosters understanding and reduces prejudice, particularly when such interactions are perceived as equal and cooperative.
Additionally, group competition and resource scarcity intensify intergroup bias. Sherif et al. (Reference Sherif, Harvey, White, Hood and Carolyn Sherif1988) contend that groups competing for limited resources tend to develop negative stereotypes and hostility toward one another. Similarly, Tajfel and Turner (Reference Tajfel, Turner, William and Worchel1979) explain that competitive situations often lead groups to devalue out-groups. Political and economic inequality further exacerbate these dynamics by fostering resentment and mistrust. Wilkinson and Pickett (Reference Wilkinson and Pickett2010) argue that disparities in wealth and power heighten prejudice, as marginalized groups may feel hostility toward more privileged groups. According to Campbell (Reference Campbell2015), this effect becomes even more pronounced during times of crisis.
Finally, education and cognitive styles significantly influence intergroup relations. Higher levels of education are generally associated with lower levels of prejudice, as education promotes critical thinking and empathy. Moreover, individuals with rigid cognitive styles are more prone to endorsing stereotypes, whereas those with flexible thinking are more likely to engage positively with diverse groups (Carney et al., Reference Carney, Jost, Gosling and Potter2008).
We acknowledge that the absence of a truly random sample may introduce biases into our analysis, including potential self-selection. However, we believe this is mitigated by the neutral incentives offered to the potential participants, namely offering a safe place for them to share their opinions, so that we reduced the chance of systematically attracting different segments of both subsamples.
Despite this, we must also acknowledge three key challenges with our methodology. First, it is difficult to accurately measure an ex-combatant’s political beliefs using a brief set of questions. Given the testing and validation of the attitude measures in previous studies, we are confident in their reliability. Yet since there is always a possibility that our survey design influenced respondents’ answers, future endeavors may attempt to include additional strategies of measurement.
Second, our reliance on veterans and ex-combatants in the process of reintegration to civil life means we cannot be certain that they fully represent the beliefs of active combatants. A larger, randomly selected sample would have strengthened our results, but accessing active combatants presents logistic and security challenges. Nevertheless, we are confident that our research identifies key relationships that shed light on the modern Colombian conflict.
Finally, our models treat ideology as static, even though we recognize it is a dynamic element. Further research is needed to explore how ideology evolves over time and how such evolution might affect the observed intergroup biases.
The manuscript’s arguments and findings align with social science studies in Latin America, a region that has experienced multiple armed conflicts and political violence. The findings from Colombia offer valuable insights for similar processes in other Latin American countries affected by violence. In particular, the text underscores the need for effective policies to address ideological radicalization, aligning with research on peace and reconciliation strategies.
Beyond armed conflicts, ideological and social polarization is a widespread phenomenon in Latin America, especially amid ongoing political and social crises. The manuscript’s findings should contribute to the literature on polarization in the region by highlighting how extreme ideologies and the legacies of past conflicts perpetuate social division.
Addressing ideological radicalization should be crucial for reconciliation, which has implications for peace-building initiatives. This perspective can inform policies in other countries seeking to consolidate peace and improve intergroup relations after violent conflict. Depolarization initiatives, aimed at improving intergroup attitudes, could complement some other efforts destined to heal individuals and communities from past harm, and perhaps mitigate some unintended negative consequences.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/lar.2025.10093
Data collection and availability
This article builds on data collected under project MinCiencias 495–2020. The study adheres to the ethical principles of human subjects research in social science. Data collection procedures were reviewed and approved by Universidad del Rosario’s Research Ethics Committee (Minute DVO005-063-CS048, February 8, 2018). Participants were fully informed about the study’s objectives, their voluntary participation, and their right to withdraw at any time. No financial compensation was provided, and informed consent was obtained before participation. Special attention was given to ensuring the confidentiality and anonymity of respondents, particularly given the sensitive nature of their backgrounds as former combatants. We express our gratitude to all former members of the Colombian Armed Forces and ex-FARC combatants for their participation in this study. A replication dataset is available at https://doi.org/10.34848/8YXUSO.
Open access
