4 Intergroup Contact Theory
written by Vanessa Müller (original draft), Milica Ninković (revision), Raul Szekely (revision), and Lukas Wallrich (revision)
4.1 The Classic
In the mid-20th century, after the horrors of World War II and during fights against official racial segregation, social scientists began asking a deceptively simple question: If you bring members of conflicting groups into contact, will they start to get along? Opinions were divided. Some warned that interracial contact would only breed “suspicion, fear, resentment, disturbance, and at times open conflict” (Baker 1934, pg. 120; cited in Pettigrew and Tropp 2006; Brophy 1945). Others were more optimistic, suggesting that isolation allowed prejudice to “grow like a disease” (Brameld 1946, pg. 245; cited in Pettigrew and Tropp 2006) and that, under the right circumstances, interaction could lead to “mutual understanding and regard” (Lett 1945, pg. 35; cited in Pettigrew and Tropp 2006). This debate set the stage for one of social psychology’s most influential ideas: intergroup contact theory.
One of the first real-world tests of these ideas came in the United States Merchant Marine shortly after World War II. In 1946, sociologist Norman Brophy surveyed white sailors now serving in newly desegregated ship crews (Brophy 1945). He created a “prejudice index” from interview questions and looked for patterns. Expected predictors of racial attitudes – such as where a sailor was born or how much education he had – turned out not to matter much. Instead, direct personal contact was the standout factor. Brophy found that white seamen who had never shipped with a Black crewmate scored highest in prejudice, whereas those who had taken four or more voyages with Black crewmates scored the lowest. In the cramped, cooperative environment of a ship – an “artificial society” where survival depended on teamwork – many sailors discovered they could no longer “afford the luxury” of prejudice. And Brophy wasn’t alone in this observation. Similar studies, mostly in the United States, showed more positive attitudes among White police officers who worked with Black colleagues (Kephart 1957), and White residents who lived in mixed buildings where they had the opportunity to interact with Black neighbours (Deutsch and Collins 1951). These early findings suggested that prejudice was not immutable, but could change with contact.
#definition Prejudice
A negative attitude toward a group and its members, often based on stereotypes rather than direct experience.
These patterns spurred social scientists to theorise why and when contact might reduce prejudice. In his landmark book The Nature of Prejudice (1954), the psychologist Gordon Allport proposed what has become known as the contact hypothesis: the idea that under appropriate conditions, interpersonal contact between members of different groups can be one of the most effective ways to reduce intergroup prejudice. Crucially, Allport (1954) did not claim that contact always works. Instead, he specified four optimal conditions that, in his view, were needed for contact to reduce prejudice:
- Equal Status: The groups should have equal status within the contact situation.
- Common Goals: The groups should strive towards a mutually beneficial outcome.
- Cooperation (Not Competition): The interaction should require cooperative effort from members of different groups.
- Support of Authorities or Norms: The contact experience should have the explicit or implicit support of authorities, law, or social norms (e.g., teachers who encourage intergroup exchange explicitly).
Allport (1954) hypothesised that when these conditions are met, contact encourages people to view one another as individuals and teammates, and thus perceive members of the “other” group as part of a shared “us” rather than a separate “them.” Interpersonal contact could then reduce ignorance and anxiety, increase empathy and understanding, and ultimately chip away at prejudice (Allport 1954). On the other hand, Allport (1954) warned that contact in unfavourable circumstances could backfire.
#yourturn
Think about a common intergroup contact situation in your community (for example, students from different backgrounds meeting at university, or neighbours from different ethnic groups interacting). Does that situation meet Allport’s four optimal conditions (equal status, common goals, cooperation, and supportive norms)? How might the presence or absence of these conditions be influencing how well the groups get along?
4.2 The Aftermath
Allport’s (1954) formulation of the contact hypothesis was hugely influential. It inspired a wave of research from the 1950s onward as psychologists, sociologists, and others researched the power of contact in a variety of groups and settings, mostly in observational research. By the turn of the 21st century, the evidence base had become enormous – though somewhat scattered. Hundreds of studies across dozens of countries and intergroup contexts had examined intergroup contact in one form or another, and the contact hypothesis had become a cornerstone of social psychology. The overarching question remained: Does contact typically work to reduce prejudice, and under what conditions?
#definition Observational Research
A study design where researchers measure variables as they naturally occur, without manipulating them. Observational studies can reveal associations between variables but cannot, on their own, establish that one causes the other.
4.2.1 The Classic Meta-Analysis
By the early 2000s, it was challenging to see the big picture in contact research. To address this, psychologists Thomas Pettigrew and Linda Tropp (2006) conducted a landmark quantitative review. In 2006, they published a meta-analysis synthesising findings from 515 studies (covering 713 independent samples and over 250,000 participants) that had studied intergroup contact. Across this vast body of work, they found a consistent pattern: people who reported more positive contact with members of an outgroup also tended to report lower levels of prejudice toward that group.
#definition Meta-analysis
A statistical technique that combines the results of multiple independent studies to estimate an overall effect. Meta-analyses can reveal patterns across a large body of research, but the quality of their conclusions depends on the quality and comparability of the included studies.
Pettigrew and Tropp (2006) concluded that “intergroup contact can promote reductions in prejudice” (p. 751) and that “there is little need to demonstrate further contact’s general ability to lessen prejudice” (p. 766), even in situations when not all optimal conditions are met. The average effect size was substantial by social science standards (Cohen’s d ≈ 0.43). With this uplifting message, their meta-analysis has become one of the most-cited papers in social psychology, with over 13,000 citations to date.
However, most of the studies they synthesised were observational rather than experimental, meaning they measured naturally occurring contact rather than manipulating it. While observational studies are valuable for spotting consistent relationships, they cannot, on their own, establish that contact caused the reduction in prejudice. For that, experiments are usually needed, and only 5% of the studies in the meta-analysis are true experiments. Pettigrew and Tropp (2006) acknowledged this limitation but advanced various arguments why their results still indicate causal effects. Most importantly, studies that used more rigorous methods (for example, longitudinal designs or experiments) tended to find larger effects of contact than weaker, correlational studies did.
Pettigrew and Tropp (2006) also aimed to assess whether the benefits of contact generalise – that is, does having a positive experience with, say, one Black teammate make a white person feel more positively toward Black people in general? Encouragingly, many studies did find evidence of generalisation: improved attitudes often extended beyond the specific individuals involved to the outgroup as a whole. For example, if a white student befriended a Latino roommate, not only might their attitude toward that roommate improve, but their overall attitude toward Latinos could become more favourable as well. This kind of generalisation is crucial if contact is to have a broad social impact, and the meta-analysis indicated that it often occurs.
For a time, Pettigrew and Tropp’s (2006) comprehensive review seemed to settle the debate: Intergroup contact works. With so many studies and an authoritative meta-analysis affirming that contact typically reduces prejudice (even outside of perfect conditions), the contact hypothesis gained even more prominence. Textbooks began to state confidently that positive contact is a proven method to improve intergroup relations. However, the story didn’t end there. Sceptics and careful scientists raised important questions and cautions that would spark the next wave of investigations. Most importantly: is the evidence causal? If we observe that people who have more friends from other groups also show lower prejudice, it’s not always clear which way the arrow of causality points – does contact reduce prejudice, or do less-prejudiced people simply seek out more contact? Pettigrew and Tropp’s (2006) analysis went a long way toward addressing this by showing that the best studies (including experiments) found stronger effects, but still, the bulk of studies in their database were not true experiments. Additionally, critics wondered about unpublished null findings: were there “file drawer” studies where contact had no effect that were never known, potentially making the published literature look overly rosy? These cautions set the stage for a new generation of research that aimed to more robustly test when and how contact works – and to probe its limits.
#yourturn
Why is it important to go beyond correlational evidence (where we simply observe relationships) when evaluating whether intergroup contact truly reduces prejudice? What kinds of studies or methods would give more convincing evidence of causation?
4.2.2 New Insights and Challenges: Refining the Theory
By the 2010s, researchers began responding to these methodological concerns, bringing fresh scrutiny to the study of intergroup contact. For instance, a review by Elizabeth Paluck, Seth Green, and Donald Green (2019) specifically re-evaluated the contact hypothesis from a rigorous causal perspective. They exclusively focused on studies that met a high bar for evidence: field experiments with random assignment to a contact condition versus a control condition, and outcome measures assessed after the contact experience was concluded. Out of the thousands of contact studies conducted over the decades, Paluck, Green, and Green (2019) found only 27 experiments that fit these strict criteria up to that point. (Notably, almost two-thirds of those 27 had been published after Pettigrew and Tropp’s (2006) meta-analysis, reflecting the field’s recent push for experimental work.)
#definition Experiment
A study where researchers deliberately manipulate one or more variables and randomly assign participants to different conditions. Random assignment helps ensure the groups are similar before the intervention, so differences in outcomes are more likely to be caused by the manipulation rather than by pre-existing differences.
The good news was that, overall, the evidence from these rigorously controlled studies still supported Pettigrew and Tropp’s (2006) basic conclusion: intergroup contact “typically reduces prejudice.” In their meta-analysis of the 27 experiments, Paluck, Green, and Green (2019) found that the average effect of being randomly assigned to a positive contact experience was a reduction in prejudice levels compared to the control groups, with Cohen’s d ≈ 0.39, very similar to Pettigrew and Tropp’s (2006) result. This helps rebut the idea that the contact-prejudice link was merely a selection effect; even when people were assigned to have contact, prejudice tended to go down, on average. However, the experimental evidence also revealed some important caveats. One striking finding was that contact’s effectiveness varied considerably by context and target group. In particular, interventions aimed at reducing ethnic or racial prejudices (for example, between Israelis and Palestinians, or between white and Black Americans) tended to show weaker effects than interventions aimed at reducing prejudice toward other stigmatised groups (such as people with disabilities or members of an opposing political party). In other words, contact worked least well for some of the most historically entrenched divides like race and ethnicity. On the flip side, contact interventions addressing prejudices that might be less emotionally charged or less tied to deep-rooted group identities (for example, toward the disabled, or between fans of rival sports teams) produced relatively larger improvements on average.
Paluck and colleagues (2019 ) also highlighted critical gaps in the evidence. For example, they found an almost complete lack of field experiments focused on adult populations dealing with racial or ethnic prejudice – the context the contact hypothesis had originally been about and arguably still one of the most important areas for policy. Additionally, very few studies had systematically tested Allport’s (1954) optimal conditions by manipulating those factors to see which mattered most. The authors concluded that these gaps need to be filled before we can confidently advise policymakers to rely on contact to remedy societal prejudice. In short, their message was not “contact doesn’t work” but rather “contact can work, but we need better evidence, especially on the toughest cases and the crucial conditions, to understand how to use it most effectively.”
#yourturn
Intergroup contact seems to yield larger prejudice reductions for some kinds of group differences (for instance, attitudes toward people with disabilities) than for others (like attitudes between ethnic groups). Why do you think this might be? Consider the nature of prejudice or anxiety in each case. What factors could make prejudice based on race/ethnicity harder to change through contact compared to prejudice toward people with disabilities, and vice versa?
4.2.3 An Outstanding Modern Study: Contact on the Soccer Field in Post-ISIS Iraq
To illustrate both the strengths and limitations of intergroup contact in action, consider a modern field experiment that put Allport’s (1954) hypothesis to a challenging test. Political Scientist Salma Mousa conducted a remarkable study in post-conflict Iraq, published in 2020, to see if positive contact could help heal rifts between deeply divided religious communities (Mousa 2020). The setting was Northern Iraq in the aftermath of the ISIS terror reign. In 2014, ISIS had overrun the region, committing atrocities including the displacement of almost the entire Christian population from certain towns. By 2016, after ISIS was defeated, many displaced Christian families began returning to their hometown of Qaraqosh, a historically Christian town that had been scarred by violence. These returning Christians carried intense distrust and resentment toward the local Muslims. The Christians feared that some Muslim neighbours had been complicit with ISIS, or at least did not suffer as they had, and rumours and grievances ran rampant. In turn, Muslim residents felt unwelcome and resented the suspicions. In this tense post-ISIS context, the two groups lived segregated lives, with social contact minimal and fraught. Prejudice and fear were high on both sides.
Mousa (2020) wondered if a carefully designed contact intervention could begin to rebuild trust and coexistence in this environment. She chose a grassroots approach: recreational soccer teams. Why soccer? Importantly, soccer in this context naturally met many of Allport’s (1954) optimal conditions for positive contact. For one, players on a team share a common goal – to win matches – and must cooperate closely to do so (passing the ball, strategising, etc.). Team sports also tend to equalise status; when everyone puts on the same jersey, they have equal status as teammates on the field. Additionally, Mousa (2020) worked with local organisations and community leaders (including church officials) to support and endorse the league, lending authority approval to this intergroup activity. In short, the intervention was deliberately structured to tick all of Allport’s (1954) boxes.
Here’s how the experiment worked. Mousa (2020) invited young Christian men in Qaraqosh who were interested in playing soccer to form teams in a new reconciliation soccer league. These men formed teams mostly with friends or neighbours, so initially, all-Christian teams. The twist was that Mousa (2020) then randomly assigned half of the league’s teams to receive several Muslim players as additions to their roster (the other half of the teams remained all-Christian and served as a control group). The Muslim players were recruited from outside the town (from camps of displaced Muslims nearby) and chosen to be of similar skill level to the Christian players, so that they could genuinely contribute on the field without dominating or being token outsiders. In total, each “mixed” team got three Muslim teammates added. All teams – mixed and all-Christian alike – then played in the same 8-week amateur league, facing each other in matches. Importantly, every other aspect of the league was the same for everyone: all teams had the same equipment, schedule, and participated under the same community-endorsed conditions, with the only difference being whether your teammates included Muslims or not. This experimental setup meant that if differences emerged between players on mixed teams versus all-Christian teams, the only systematic explanation would be the experience of having (or not having) Muslim teammates.
At first, the intervention faced friction. Some Christian players were unhappy about Muslims joining their teams. In the early weeks, there were incidents of mistrust and even hostility – for example, a few Christian team members openly told the organisers “We don’t want Muslims; they will ruin the league.” Such remarks underscored just how deep the suspicion ran in this community; it wasn’t an easy start. But as the season progressed and these young men practised and competed side-by-side, the tone began to shift. By about the mid-point of the season, signs of camaraderie had emerged. One small episode stood out: when some Christian players learned that their new Muslim teammates were struggling to afford taxi fare to the games (travelling from a distant displacement camp), the Christian players pooled money to help cover the cost so their teammates could make it to matches. On the field, teammates started to celebrate goals together and encourage one another. Over time, a shared team identity – we are the Lions, we are teammates – began to form, overlaying the previous religious divide. A Christian player, asked later about his experience, reflected that “I learned that Muslims could be friends of ours, even like brothers.” The transformation was not instant or universal, but by the end of the league, many of the initial anxieties had given way to friendly competition and mutual respect on these mixed teams.
So, did this Allportian (Allport 1954) contact experience actually change attitudes or behaviors? Mousa’s (2020) results were revealing. They showed both encouraging positive outcomes and clear limits. First, consider the effects within the context of the league itself – that is, how the Christian players felt and acted toward their Muslim teammates (and other Muslims in the league): The Christian players who had Muslim teammates ended up displaying significantly more positive behaviours toward Muslim peers compared to players on all-Christian teams. For example, at the end of the season, each team voted for a member of an opposing team to receive a sportsmanship award. Christians on mixed teams were more than 15 percentage points more likely to vote for a Muslim player (from another team) for this award than were Christians on all-Christian teams. This indicated greater esteem and fairness toward Muslim peers. Moreover, when sign-ups opened for a new season, the mixed-team Christians were much more willing to play on a mixed team again – they registered at higher rates for a subsequent mixed league – whereas many all-Christian team players declined to sign up once they heard teams might be mixed. Perhaps most impressively, about six months after the experiment, Mousa (2020) found that many of the mixed-team players were still regularly meeting up with their former Muslim teammates to practice together and maintain their friendship. In fact, roughly one-third of the mixed teams continued to meet socially for pick-up soccer games long after the official league ended, whereas almost none of the all-Christian teams chose to continue gatherings that included outgroup members. These findings show that meaningful friendships and trust did form through the contact intervention. By all accounts, prejudice had decreased, at least with respect to those specific Muslim teammates and other known Muslim players.
However, now consider what happened outside the context of the league – in the broader community and in attitudes toward Muslims in general. Here, the findings were more sobering: The positive effects of contact did not substantially generalise to Muslims beyond those directly encountered. In surveys and behavioural measures after the season, Christian participants who had played with Muslim teammates showed no significant change in their willingness to interact with unknown Muslims or visit Muslim communities compared to the control group. For instance, having had Muslim teammates generally did not make Christian players more likely to say they would patronise a restaurant in a nearby majority-Muslim city, nor did it increase their attendance at a mixed social event in town. When asked about broader attitudes, those who experienced contact did express somewhat stronger abstract support for coexistence or the idea that Christians and Muslims could be friends, but their core beliefs about Muslims as a group (for example, levels of trust toward Muslim strangers or stereotypes about Muslims) remained essentially as negative as before. In Mousa’s (2020) own words, while the Christian players found it possible to trust and befriend specific Muslim individuals they got to know, extending trust to Muslim strangers outside that circle was “too much of an ask” in the aftermath of war. In short, the contact intervention succeeded in forging new cross-group friendships and improving attitudes toward those individuals, but it largely failed to shift the participants’ generalised feelings about the outgroup as a whole or their behaviour in other contexts.
This pattern – friendships without broad reconciliation – highlights a crucial challenge for intergroup contact theory. Mousa’s (2020) study offers an inspiring proof-of-concept that even in a highly fraught, post-conflict setting, a well-designed contact program, featuring Allport’s (1954) optimal conditions, can produce genuine goodwill and cooperation between former adversaries. The fact that young men who initially hated the idea of playing with “the other” ended up forming lasting bonds is powerful. It shows that under the right conditions, enemies can indeed become teammates, even friends. On the other hand, the limited scope of these changes tempers the optimism. The contact in this study changed how people felt about particular outgroup members, but not necessarily about the outgroup at large. From a policy or peacebuilding standpoint, that is a big limitation: improving one-to-one relationships is wonderful for those individuals, but it may not significantly mend the overall social fabric or reduce the kind of generalised fear that fuels wider conflict. Mousa’s (2020) findings align with what many other studies have found and what is now a central puzzle in contact research – the generalization problem. How can we ensure that the effects of contact spread beyond the immediate participants and influence attitudes more broadly? If positive contact only affects the small circle of people directly involved, its ability to reduce community-wide prejudice or conflict is limited.
#yourturn
In the soccer study, Christian players clearly grew more accepting of the Muslim teammates they got to know personally, yet their attitudes toward Muslim strangers remained unchanged. Why do you think a positive experience with a few individuals might fail to generalise to the entire outgroup? What psychological factors might be at play? Can you think of any additional measures or tweaks to the intervention that might help encourage broader changes in attitudes or trust (for example, activities that mix the groups in other settings, discussions that address group stereotypes, etc.) to help bridge that gap?
Mousa’s (2020) soccer experiment encapsulates both the promise and the limitations of intergroup contact. It provides a vivid example that contact can work – even under pretty challenging conditions, it built trust and friendship where there was initially fear and hostility. At the same time, it underscores that a single intervention, even a well-crafted one, is no panacea for deeply rooted prejudices. Especially in contexts of recent violence and trauma, biases may run so deep that it takes much more than a brief intervention to budge generalised attitudes. These nuanced outcomes have prompted researchers to investigate strategies to amplify and extend contact effects. How might we design contact interventions that not only improve attitudes toward the people directly involved, but also shift perceptions of the broader group? This remains an active area of research.
In fact, as the field has progressed, experts have adopted a more cautious tone about what contact can realistically achieve on its own. In 2021, Paluck et al. (2021) published an extensive review of 418 prejudice-reduction experiments conducted between 2007 and 2019, a collection that included many contact-based interventions alongside other approaches. The results of this review were mixed and somewhat concerning. On one hand, many of the experiments reported at least some positive effects on attitudes, suggesting there are reasons for optimism. On the other hand, the authors uncovered “troubling indications of publication bias,” meaning that studies showing big success were likely overrepresented in the literature, while those with null or tiny effects may not have been published. When they statistically accounted for this bias, the overall picture became less rosy. Furthermore, three-quarters of interventions in that review were very “light-touch” or brief, such as a short workshop, a single encounter, or a one-time media exposure. Not surprisingly, any positive changes from such brief interventions often faded over time or were quite limited in scope. In the relatively few cases where more intensive, long-term interventions were implemented (what the authors called “landmark studies”), the effects on prejudice tended to be modest at best. This included some multi-week educational programs, extended intergroup dialogues, and other sustained efforts – many showed only small improvements, highlighting how stubborn prejudices can be. Paluck-et-al_2021 concluded that new theoretical innovation is needed to achieve larger and more lasting impacts. They suggested that perhaps contact on its own is often too limited, and that combining contact with other approaches (or addressing larger structural issues in tandem) might be necessary to produce more substantial change. In their view, simply throwing diverse people together for a short period is rarely a magic fix; researchers need to think bigger about the mechanisms of change and consider multi-pronged solutions.
Most recently, the strongest tests of the contact hypothesis have been compiled in a 2025 meta-analysis by economist Matt Lowe (Lowe 2025). Lowe (2025) focused exclusively on the highest-quality studies: those that were pre-registered, randomised experiments on intergroup contact.
#definition Pre-Registered Study
A study in which the researchers publicly register their hypotheses, methods, and analysis plan before collecting data. Pre-registration helps increase transparency and credibility – it prevents researchers from changing their analyses or selecting results after seeing the data, which can lead to false-positive findings.
By zeroing in on these rigorously planned studies, Lowe (2025) aimed to eliminate biases introduced by practices like p-hacking or cherry-picking of data – practices that can inflate apparent effects.
#definition p-hacking
The practice of misusing data analysis to find patterns that can be presented as statistically significant, often by trying many variable combinations or statistical tests until something “significant” turns up. This can lead to unreliable conclusions because it capitalises on chance patterns in the data.
#definition Cherry-Picking
Reporting only the data, outcomes, or time frames that support one’s hypothesis while ignoring or dismissing those that do not. This makes the story or articles simpler and might make them more publishable, but provides a distorted view of the evidence.
The findings are instructive. When considering only these methodologically pristine studies, the average effect of intergroup contact on prejudice outcomes was much smaller than earlier reviews had suggested, with d ≈ 0.1, a quarter of the effect size suggested by Pettigrew and Tropp (2006). In plain language, this means that the effect is statistically significant, but on average, quite modest. Lowe’s (2025) meta-analysis also reinforces the now-familiar theme about specificity vs. generality: contact’s benefits tend to be localised. People’s attitudes and behaviours toward the particular individuals they met often improved more strongly than their attitudes toward the outgroup in general. Broad attitude change was much less common, with many studies finding little to no shift in generalised prejudice or policy views even when interpersonal warmth increased. That contact’s effects often fail to generalise widely is now recognised as one of the central challenges in the field.
4.3 Conclusion
Seven decades after Allport (1954) first set out the contact hypothesis, it remains a cornerstone of prejudice‐reduction research. From post‐war merchant ships and integrated housing projects to modern field experiments, the idea has consistently shaped both science and policy. Its core message that prejudice is not fixed and can change through structured, positive interaction helped shift thinking away from segregation toward integration as a deliberate tool for improving relations.
The evidence, however, shows that contact is no cure‐all. Gains are often local, improving attitudes toward specific individuals but failing to generalise to the wider group. Outcomes depend heavily on context, structure, and the quality of interaction. For that, Allport’s (1954) optimal conditions (equal status, common goals, cooperation, and authority support) remain a useful guide, though further research is needed. Contact is also not always positive, and researchers have started taking that more seriously, as negative experiences can be as powerful, if not more so, than positive ones, though they are fortunately rare (Paolini et al. 2024). Contact also needs to be understood in context, as broader forces such as inequality, political division, and historical grievances can limit its impact. Here, psychologists can fruitfully cooperate with other social science disciplines.
Studying the most meaningful forms of contact – deep, sustained relationships forged over years – poses particular challenges. Such relationships cannot be randomly assigned, develop slowly, and are difficult to measure without disrupting them. Creative, flexible designs are therefore needed, with interpretations that acknowledge their limitations. Current research focuses on making contact more effective and lasting. Promising approaches include pairing it with perspective‐taking, cooperative learning in schools, norm‐shaping media campaigns, or virtual‐reality simulations of positive encounters. Some initiatives also embed contact in long‐term community projects or redesign institutions, such as integrated workplaces or mentoring networks, so diverse cooperation becomes part of daily life. The challenge ahead is to move from showing that contact can work to understanding how to make it work consistently, at scale, and for the long term.
#yourturn
The IAT was designed to assess automatic associations people may hold unconsciously. However, if implicit and explicit attitudes are highly correlated, what are the implications for how we understand the relationship between conscious and unconscious mental processes?