When the lens is too wide: The political consequences of the visual dehumanization of refugees

The paper investigates how dominant visual framings in photojournalism influence public attitudes and political behavior toward refugees. While iconic images of identifiable victims can elicit empathy and prosocial responses, Western media more commonly show refugees as large, faceless groups, especially during the Syrian refugee crisis. The central research questions are whether, and how, the prevalent framing of refugees in large groups (versus small, more identifiable groups) dehumanizes them, and whether such dehumanization carries political consequences. The authors hypothesize that images depicting large groups will increase implicit dehumanization (reduced attribution of uniquely human, secondary emotions), alter explicit perceptions of humanness, reduce support for pro-refugee policies, increase support for anti-refugee policies, and shift leader preferences toward more authoritarian-looking candidates. They also test whether narrative context (sea vs. land depictions of refugee journeys) modulates these effects and explore emotional mechanisms (pity, admiration, contempt) underlying changes in political choices and perceived threat.

The study builds on several literatures: (1) Media studies and political communication emphasize the political power of images and framing effects in shaping public understanding and responses to crises; (2) Social psychology documents the identifiable victim effect, where identifiable individuals elicit greater empathy and aid than statistics or large groups, and psychic numbing in response to mass suffering; (3) Dehumanization research shows outgroups are often denied uniquely human traits or moral worth, linked to negative attitudes and policy preferences; (4) Work on the stereotype content model relates reduced warmth and competence to dehumanizing perceptions; (5) Studies of refugee media coverage find dominant portrayals as large, anonymous groups (often via medium/long shots), sometimes embedded in metaphors of flooding or security threats, potentially emphasizing threat over humanitarian narratives. The authors highlight gaps concerning the political consequences of the dominant large-group framing, and the role of contextual narratives (e.g., sea crossings) in modulating dehumanization and downstream political behavior.

Design and participants: Across 10 studies (8 main plus 2 supplementary) with 3,951 European participants recruited via Prolific, the authors used online experiments (Qualtrics). Studies were not preregistered. Ethical approval obtained; informed consent collected; attention checks included. Participants were compensated ~£7/h. Stimuli: Award-winning photojournalistic images of refugees (1990–2017) were coded into two framings: small groups (pSG; recognizable faces) versus large groups (pLG; no clearly recognizable faces). Final set: 43 pSG and 39 pLG images. For Studies 2 and 8, images were further categorized by narrative context: sea (on boats/near sea) versus land. Measures and tasks: - Dehumanization (implicit): Attribution of primary vs. secondary emotions to refugees (pre- and post-exposure) on VAS; primary emotions serve as controls; changes in secondary emotions index implicit dehumanization. - Warmth and competence (SCM): After each image, ratings of perceived warmth and competence. - Explicit humanness: Ratings of how humanely images (Study 4), headlines (Study 5), and image–headline front covers (Study 5) portray refugees. - Political behavior: Petition-signing task for pro-refugee and anti-refugee policies (Study 6); leader choice task selecting between faces varying in dominance and trustworthiness (Study 7 pre/post, Study 8 pre/post). - Emotions experienced toward refugees: VAS ratings of pity/sympathy (pity), admiration, and contempt (composite of negative affect items) during exposure (Study 8). - Additional covariates: Social dominance orientation (SDO), political orientation, xenophobia scale (fear of immigrants/foreigners), perceived realistic and symbolic threat. - Distress ratings after each image. Design specifics: - Study 1: Between-subjects pLG vs. pSG; pre–post secondary emotion attribution. - Study 2: 2×2 between-subjects Framing (pLG/pSG) × Narrative (Sea/Land) with pre–post dehumanization. - Study 3: Within-subjects blocks of refugee and survivors of natural disasters (SND), each with pLG and pSG; warmth/competence outcomes; mixed models. - Study 4: Explicit humanness ratings of images (balanced subsets of pSG/pLG); tests Framing and Narrative. - Study 5: 2×2 within designs pairing headlines (humanizing vs. dehumanizing) with photos (pSG vs. pLG) into front covers; explicit humanness ratings; mixed models. - Study 6: Petition-signing task after exposure to pLG or pSG; ordinal logistic regressions controlling for SDO, political orientation, dehumanization change, distress. - Study 7: Leader choice task (pre/post) with faces varying in dominance/trustworthiness; tests whether pLG exposure increases preference for dominant/less trustworthy leaders; models include xenophobia and other covariates. - Study 8: Leader choice task (pre/post) plus emotion ratings during exposure; perceived symbolic/realistic threat; structural equation modeling (SEM) to test mediation via pity, admiration, contempt for Framing and Narrative effects. Analyses: Linear/mixed-effects regressions (R packages lme4, car, emmeans), ordinal logistic regressions for petitions, logistic regressions for leader-choice probabilities, SEM (lavaan), and a fixed-effect meta-analysis (metafor) on dehumanization across Studies 1, 2, 6, 7. Covariates were centered/scaled; planned post-hoc comparisons with corrections.

- Implicit dehumanization: Exposure to pLG (large, faceless groups) reduced attribution of secondary, uniquely human emotions to refugees compared to pSG (small, identifiable groups) (Study 1; e.g., model R² ≈ 0.689; Framing effect F(1,640)=4.42, p=.036). - Narrative moderation: In Study 2, the Framing × Narrative interaction was significant (F≈6.52, p=.010). The pLG vs. pSG dehumanization difference was present for sea images (t≈2.90, p<.001) but not for land images. - SCM replication: Study 3 showed a Framing × Group interaction for warmth (χ²=13.38, p<.001) and competence (χ²=20.06, p<.001): both refugees and SND were judged warmer after pSG vs. pLG, with a stronger warmth reduction for refugees in pLG. - Explicit humanness: Study 4 found images of large groups (pLG) were rated as less humane than pSG (p<.001). Sea images were, on average, rated more humane than land images (p≈.0025); no significant Framing × Narrative interaction. - Image–text combinations: Study 5 showed both dehumanizing headlines and pLG independently lowered perceived humanness of front covers, with combinations of dehumanizing headlines and pLG yielding the lowest humanness ratings. - Policy support: Study 6 showed that exposure to pLG decreased support for a pro-refugee petition (t≈-2.51, p≈.012) and increased support for an anti-refugee petition (t≈2.39, p≈.017), controlling for SDO, political orientation, dehumanization change, and distress. - Leader preferences: Study 7 found exposure to pLG (vs. pSG) increased the probability of choosing a more dominant and less trustworthy-looking leader (R²≈0.015; F≈3.97, p≈.047). The effect was stronger among participants higher in xenophobia (interaction p≈.008). - Meta-analysis: Across Studies 1, 2, 6, and 7, a fixed-effect meta-analysis confirmed greater dehumanization after pLG vs. pSG on average (z≈-2.26, p≈.002), despite non-significant dehumanization effects in Studies 6 and 7 individually. - Mechanisms (Study 8): SEM revealed that pLG exposure reduced pity and admiration and increased contempt; lower pity mediated the effect of Framing and Narrative on increased preference for authoritarian-looking leaders. Framing and Narrative also influenced perceived symbolic (and to a lesser extent realistic) threat via reduced pity/admiration and increased contempt.

The studies collectively demonstrate that the dominant media portrayal of refugees as large, faceless groups fosters dehumanization and shifts political attitudes and choices. Framing refugees in large groups reduces attributions of their uniquely human emotional capacities and lowers perceived warmth and explicit humanness. These dehumanizing perceptions translate into meaningful political consequences: reduced support for pro-refugee policies, increased support for punitive measures, and a shift toward more authoritarian-looking leader preferences. The narrative context matters: sea depictions amplify dehumanization effects, consistent with broader media metaphors that frame refugees as elemental forces or security threats. Mechanistically, emotions experienced by viewers—particularly reduced pity—mediate the link between visual framing and political choices and threat perceptions, indicating that images shape socio-political evaluations through affective pathways rather than solely through changes in attributed emotions to the outgroup. These findings advance understanding of how visual media can independently and powerfully shape public opinion and democratic outcomes.

This work integrates insights from social psychology, media studies, and the humanities to show that common photojournalistic framings—especially large, anonymous depictions of refugees—have measurable social and political effects. Across 10 studies, the authors demonstrate that such framings heighten dehumanization, reduce explicit humanness judgments, decrease support for pro-refugee policies, increase support for anti-refugee measures, and bias leader preferences toward more dominant and less trustworthy-looking candidates. Emotional reactions, particularly reduced pity, help explain how framing and narrative context influence political behavior and perceived threat. Future research should examine additional visual dimensions (e.g., iconography, emotional expressions, camera distance), explicit forms of dehumanization, cross-national differences tied to geopolitical contexts, and the dynamics of image–text interactions in real-world media ecologies. These efforts can inform ethical editorial choices and public communication strategies that mitigate dehumanization and its political consequences.

- Non-preregistered studies may raise concerns about analytic flexibility. - Mapping of effects varied across studies: dehumanization effects were not significant in Studies 6 and 7, though confirmed by meta-analysis; differences in design (e.g., inclusion of both sea and land images, task ordering) may account for variability. - Online, self-report measures and brief exposures may not fully capture the complexity of real-world media consumption. - European sample limits generalizability to other regions and cultural contexts; potential country-level moderators were not systematically analyzed. - Focus on group size and facial identifiability (pLG vs. pSG) and sea/land narrative does not encompass other influential image features (e.g., emotion expressions, composition, iconography). - Explicit dehumanization was not the primary focus; future work should examine its relation to visual framing and political outcomes.