Human large-scale cooperation as a product of competition between cultural groups

The study addresses why humans frequently cooperate with unrelated, unfamiliar individuals in transient interactions—a pattern not well explained by classical evolutionary theories of cooperation. Cultural Group Selection (CGS) posits that culturally structured population differences, maintained despite migration, permit group-level selection on cultural norms that benefit groups. A key prediction is that the social scale of cooperation should correspond to the scale and magnitude of cultural differentiation: greater cultural differentiation between groups at a given level should be associated with more parochial norms limiting cooperation across that boundary. Using cultural FST (the proportion of total cultural variation between groups) as the critical statistic for between-group selection strength, the authors hypothesize a negative association between cultural FST and cooperation across groups. They test this in pastoralist societies in northern Kenya (Turkana, Samburu, Rendille, Borana), where intergroup competition over resources and raiding is salient and social organization is nested (clans, territorial sections, ethnic groups).

CGS has been widely discussed but insufficiently tested directly. Prior support includes historical/ethnographic cases, arguments that preconditions for CGS are met in humans, and demonstrations that institutions can be group-beneficial. Bell et al. (2009) showed substantial cultural FST between countries, implying scope for CGS, but did not relate it to cooperation patterns. Francois et al. (2018) linked increased competition among firms to increased within-firm cooperation but did not analyze nested population structures. Cultural FST values reported in prior work (e.g., World Values Survey, folktales, songs, Hadza contributions) are generally an order of magnitude higher than genetic FST across human groups, supporting the plausibility of CGS. Competing accounts include misfiring of reciprocity psychology based on familiarity/proximity, coordination-based explanations for cooperation with culturally similar others, and within-group norm evolution shaped by individual benefits or elite interests. The authors argue these alternatives struggle to explain the specific social scale at which cooperation aligns with cultural differentiation observed here, particularly regarding warfare norms.

Design and sampling: The study sampled 793 individuals (759 completed cultural norms; all but four completed 16 cooperation vignettes) across 9 clans within four ethnolinguistic groups (Borana, Rendille, Samburu, Turkana) and, for Turkana, 3 territorial sections (Ngikwatela, Ngiyapakuno, Ngibochoros). Sampling aimed to balance gender and ages, with one respondent per household. Fieldwork spanned April 2015–July 2016 across settlements identified based on migration/settlement patterns, access, and security. Survey development: Drawing on extensive ethnographic experience, the team generated an initial pool of 170 normative items across five domains (cooperation/helping, crime/punishment, raiding, cultural markers, family dynamics) and 26 cooperation vignette scenarios, piloted in one Turkana location (Ngiyapakuno). After piloting and back-translation refinements, the final instruments included 49 normative items (10 cooperation, 9 crime/punishment, 9 raiding, 10 family, 11 cultural markers) and 16 vignettes. Items were phrased as agree/disagree statements, with varied affirmative/negative wording to minimize acquiescence bias. Instruments were translated by local assistants and refined via iterative back-translation. Measurement of cultural differentiation: For each of the 49 norms, the frequency of endorsement was computed per group (clan, territorial section, ethnic group). Pairwise cultural FST for each norm and average FST across norms were calculated between: pairs of clans within each ethnolinguistic group; pairs of Turkana territorial sections; and pairs of ethnolinguistic groups. FST was computed as the ratio of between-group to total variance in each trait, weighted by sample sizes: FST = [(n_i p_i^2 + n_j p_j^2) − p^2] / [p(1−p)], where p is the overall frequency across groups. Measurement of cooperation: Sixteen vignettes depicted a protagonist able to help or harm an unknown target individual. Twelve scenarios involved helping (e.g., sharing water), four involved refraining from harm (e.g., not stealing livestock). The target’s group identity was experimentally varied: same clan; different clan within same ethnic group; different ethnolinguistic group (unspecified neighbor); for Turkana additional conditions of same/different territorial section; for Rendille an additional condition where the target was explicitly Samburu (given atypically friendly Rendille–Samburu relations). Each participant was assigned one condition and judged each scenario’s cooperative action (right/wrong). The cooperation rate is the proportion endorsing helping/not harming across the specified boundary. Geographic distance: GPS coordinates recorded at interview locations were used to compute mean pairwise distances between all subjects in two groups; this average served as the geographic distance between groups. Statistical analysis: A mixed-effects logistic regression (lme4::glmer in R 3.6.0) modeled the binary endorsement of cooperation as a function of fixed effects: cultural FST between actor’s and target’s groups and geographic distance between groups. Random effects: subject ID, vignette scenario, and subject’s lowest-level group membership (e.g., specific clan/territorial section). For each subject-condition, FST and distance were assigned as averages over the relevant pairwise combinations (e.g., between sampled clans for “different clan” conditions, between ethnic groups for “different ethnic group”). Model fit reported marginal and conditional R-squared (MuMIn::r.squaredGLMM). Additional models used FST computed within each domain (cooperation, crime/punishment, raiding, cultural markers, family) to assess which domains’ cultural differentiation most strongly predict cooperation. Ethics and data management: Informed consent obtained; protocol approved by Arizona State University IRB. Data collection digitized via Open Data Kit (ONA platform). Data and code are available on OSF (https://doi.org/10.17605/OSF.IO/HRJK7).

- Cultural differentiation levels: Average cultural FST between clans and between Turkana territorial sections ranged from 0.002 to 0.058; between ethnolinguistic groups from 0.087 to 0.215. Thus, up to ~20% of trait variation lies between ethnic groups, indicating substantial structure for potential CGS. - Cultural FST vs. cooperation: Cultural FST had a strong negative effect on cooperative norm endorsement across group boundaries (log-odds ≈ −20.12, p < .001). Increasing FST from 0.05 to 0.15 nearly halved the predicted probability of endorsing the cooperative act. Geographic distance had no significant effect (log-odds ≈ 0.14, p < .1). Model R2: marginal ≈ 0.18; conditional ≈ 0.57. - Domain-specific effects: When predicting cooperation with FST computed within domains, FST based on raiding norms had the largest negative effect (log-odds ≈ −19.28), followed by cultural markers (log-odds ≈ −6.02; both significant). Other domains were not significant predictors. This suggests warfare-related norms most strongly align the social scale of cooperation with cultural differentiation. - Correlation with distance: Cultural FST and geographic distance were highly correlated (Pearson r ≈ 0.93, p < .001), strongest for cultural markers and cooperation (r ≈ 0.9) and lowest for raiding (r ≈ 0.63). Given raiding FST best predicts cooperation and is least correlated with distance, the FST–cooperation association is unlikely to be driven by distance per se. - Subpopulation patterns: The negative FST–cooperation relationship held across all 15 subpopulations. Borana showed the highest cross-ethnic cooperation; Islamic influence may extend cooperative norms. Rendille exhibited unusually high cooperation with Samburu despite greater cultural similarity to other Rendille, consistent with strategic alliances. - Scenario patterns: The FST–cooperation association was consistent across all 16 scenarios. Highest cooperation and least decline with FST occurred for aiding vulnerable targets (e.g., sharing relief food; helping an injured person reach a hospital). Lowest cooperation involved high-trust transactions (entrusting a stranger to sell a cow), suggesting CGS may more strongly shape norms for transient, lower-monitoring contexts.

Findings confirm a central CGS prediction: the social scale of cooperative norms corresponds to the scale of cultural differentiation. Cultural FST strongly and negatively predicts cooperation across group boundaries, whereas geographic distance does not, supporting the view that group-structured cultural variation, not mere proximity or familiarity, shapes cooperative norms. The especially strong predictive power of raiding norms indicates that intergroup competition over critical resources selects for norms directing high-stakes cooperation at the relevant social scale. Alternative theories face challenges: a misfiring-familiarity account would predict effects of geographic proximity, which were not observed. Pure coordination-based explanations do not address costly helping/not-harming where defection yields private benefits without sanctions. Within-group norm evolution driven by self-interest or elite interests does not readily explain why norms channel cooperation along clan/ethnic lines despite apparent gains from broader mutual restraint (e.g., raiding). Contextual cases (Borana’s cross-ethnic generosity possibly via religion; Rendille–Samburu alliance norms via elder deliberation) suggest that other norm-evolution processes can operate, but the overarching alignment between cultural differentiation and cooperation supports CGS as a key driver. Implications extend beyond the studied populations: In acephalous, competitive pastoralist settings akin to much of human prehistory, CGS likely shaped cooperative psychology and institutions. The framework may also elucidate patterns in violence, morality, and religion, and inform models of social change in multicultural states (immigration, acculturation, extension of support).

The study provides direct evidence that the social scale of cooperative norms tracks the population structure of cultural variation, supporting Cultural Group Selection as a mechanism shaping human large-scale cooperation. Cultural FST robustly predicts cooperation across group boundaries, particularly for warfare-related norms, while geographic distance does not. These results imply that CGS has likely influenced human cooperative psychology and contemporary norm dynamics and helps explain humans’ combination of broad cooperative potential and cultural parochialism. Future research should: test CGS predictions across additional social scales (e.g., religions, nations) and societies; disentangle causal pathways linking cultural differentiation, competition, and norm content; integrate behavioral measures beyond normative endorsement; and model how immigration and acculturation reshape cultural structure and cooperation in modern polities.

- Causal attribution: The data cannot disentangle whether cooperative norms are directly produced by CGS or arise via within-group norm processes built on psychologies shaped by past CGS. - Measurement scope: Cooperation was measured as normative endorsement in vignettes rather than observed behavior; generalizability to actions may vary by context. - Correlation with distance: Cultural FST correlates strongly with geographic distance, though analyses suggest distance is not the primary driver; residual confounding by unmeasured factors linked to distance cannot be fully ruled out. - Context specificity: Findings derive from four pastoralist ethnolinguistic groups in northern Kenya; patterns may differ in other ecological, political, or institutional settings. - Special cases: Exceptions (e.g., Rendille–Samburu alliance norms) indicate that deliberate institutional processes can shape norms independent of baseline cultural differentiation, potentially moderating the general pattern.