Environmental variability supports chimpanzee behavioural diversity

The study investigates whether environmental variability—across short-, mid-, and long-term timescales—promotes within-species behavioural diversity in chimpanzees. Building on the premise that larger brains and innovation correlate with adaptability to variable environments in birds and primates, and that environmental instability influenced hominin evolution, the authors test if ecological variability shapes the distribution of diverse behaviours (including cultural traits) within a single species. This addresses the gap in linking environmental variability directly to intraspecific behavioural diversity, providing comparative insight relevant to hypotheses about human evolution (e.g., variability selection vs. habitat-specific hypotheses).

Prior work links brain size and innovation propensity across taxa to environmental variability and survival in seasonal habitats. Evidence across primates for innovation rates correlating with climatic variability is mixed, potentially due to failure to capture historical processes. Hominin evolutionary hypotheses emphasize either shifts from forest to savannah or selection by environmental unpredictability (variability selection). Behavioural flexibility and cultural capacities have been critical in human evolution, but fossil records limit direct tests. A comparative approach using nonhuman great apes is warranted. Previous chimpanzee work (e.g., Lycett et al. 2009) examined rainfall and cultural behaviour distribution with mixed outcomes using cladistics. The authors note chimpanzees’ wide geographic range and rich population-level behavioural variation, including cultural tool use, making them ideal for testing environmental influences on behavioural diversity.

Design: Observational comparative study of 144 wild chimpanzee communities across Equatorial Africa. For each community, the occurrence (1/0) of 31 non-universal behaviours (many cultural) was coded from direct (camera traps or observers) and indirect evidence (artefacts, feeding remains, faeces). Behaviours span extractive foraging (e.g., algae fishing, ant-dipping, pestle pounding, nut cracking, tool-assisted hunting), water extraction, thermoregulation (e.g., cave use, bathing), and communication, and were grouped into 13 behavioural categories; behaviours were also classified as tool use, non-tool use, or unknown. Data sources: 46 communities from the Pan African Programme (PanAf) using standardized camera-trap grids (1×1 km cells; 9–143 km² coverage; observation 12–30 months for most sites), and 106 communities compiled from ~450 publications (1951–2017). Presence was coded probabilistically; lack of observation is not true absence. Behaviours constrained by unavailable local resources were coded NA. Environmental predictors (capturing variability across timescales):

• Short-term: Precipitation seasonality (WorldClim BIO15; coefficient of variation of monthly rainfall averaged 1960–1990; 1 km² resolution). Proxy for ecological variability and resource availability.
• Mid-term: Predominant habitat classification (forest vs. savannah woodland) from multiple GIS layers (WWF ecoregions, TNC ecoregions/biomes, EC Global Landcover). Mosaic sites were assigned based on literature/PanAf observations. Savannahs treated as more variable than forests.
• Long-term: Distance (km) from the nearest Pleistocene forest refugia boundary (per Maley 1996) computed from community center coordinates; zero indicates inside a refugium. Controls: Human footprint index at community center (known negative effect), observation effort (months), and chimpanzee subspecies (P. t. verus/ellioti/schweinfurthii/troglodytes). Spatial autocorrelation accounted for via a Gaussian process over latitude/longitude. Statistical analysis: Bayesian Regression Models (R, brms v2.8.0; Bernoulli response, logit link) modeled the probabilistic occurrence of behaviours per community. Due to collinearity among predictors (distance–precipitation CV r=0.720; distance–habitat r=0.660; precipitation CV–habitat r=0.613; all P<0.001), three separate models were fitted, one per environmental predictor, including the same controls. Random effects: site and behaviour; random slopes for environmental predictor, human footprint, observation months within site, and subspecies within behaviour, including correlation parameters. Predictors were square-root transformed (distance) and z-transformed (numerical) where appropriate. Default flat priors used; sensitivity with weak/wide priors checked. MCMC diagnostics indicated convergence (R-hat <1.01; no divergent transitions). Robustness checks excluded certain site subsets (e.g., not fully cleaned PanAf videos, long-term habituated sites, 1-month observation sites). Additional response definitions tested: number of behavioural categories, tool-use only, and non-tool-use only. Software and code: R 3.5.3; brms; analysis code in Supplementary Code 1.

• Main result: Chimpanzee behavioural diversity (probability of occurrence of 31 behaviours) increases with environmental variability across timescales, strongest for long-term variability.
• Distance from Pleistocene refugia: Positive effect on behaviour occurrence (estimate 0.523 ± 0.228 SD; 95% CI 0.072–0.977); posterior probability of positive association = 0.990.
• Precipitation seasonality (CV): Positive but weaker (0.314 ± 0.255; 95% CI -0.199–0.794); posterior positive = 0.888.
• Habitat type: Savannah woodland vs forest positive but weaker and less certain (0.608 ± 0.596; 95% CI -0.591–1.745); posterior positive = 0.849.
• Controls: Observation months strongly positive (0.905 ± 0.294; 95% CI 0.327–1.491; posterior positive = 0.998). Human footprint negative (-0.300 ± 0.161; 95% CI -0.630–0.004; posterior positive = 0.031), consistent with prior work.
• Alternative response metrics: • Behavioural categories (13): Positive associations for all three predictors, strongest for distance to refugia (0.473 ± 0.359; 95% CI -0.262–1.156), with posterior support 95–84% across predictors. • Tool use and non-tool use subsets: Distance to refugia consistently positive for both; precipitation seasonality and habitat showed smaller and more variable effects.
• Robustness: Results qualitatively unchanged when excluding certain site subsets; non-tool behaviours underrepresented showed more variability.
• Spatial covariance declined to near zero after ~50–70 km, indicating appropriate control for spatial autocorrelation.

Findings support that environmental variability fosters within-species behavioural diversification in chimpanzees, with the most consistent signal from long-term historical variability (distance from Pleistocene forest refugia). Short- and mid-term variability (precipitation seasonality and savannah woodland habitat) also show positive associations. As many of the behaviours are cultural, results imply that ecological variability promotes both behavioural and cultural diversity. The positive association with distance from refugia suggests dispersing populations encountering novel, more variable environments may have innovated and retained additional behaviours (potentially culturally), whereas populations persisting within stable refugia may exhibit limited diversification (consistent with the museum hypothesis) or potential loss of behaviours due to reduced selection pressure or stochasticity. The authors also discuss potential roles of savannah refugia during Pleistocene contractions and note present-day chimpanzee adaptations to both forest and savannah. Broader implications include parallels to human evolution, where fluctuating environments likely shaped adaptive behavioural repertoires; and conservation relevance, as behavioural/cultural flexibility might buffer against climate change, though human impacts are concurrently eroding behavioural diversity.

Environmental variability across recent and historical timescales is associated with greater behavioural (and cultural) diversity in chimpanzees, with the strongest and most consistent effect from historical distance to Pleistocene forest refugia. This supports the idea that fluctuating environments drive intraspecific behavioural diversification, offering comparative insights into great ape, including human, evolution. Future research directions include: integrating genome-wide data to reconstruct population history and dispersal relative to refugia; disentangling the roles of forest versus savannah refugia; quantifying social learning’s role in the spread/maintenance of behaviours and potential cultural niche construction; improving measures of environmental variability beyond habitat dichotomies; and incorporating cultural trait considerations into conservation planning under climate change.

• Collinearity among environmental predictors prevented simultaneous modeling; effects were tested in separate models.
• Behavioural occurrences are contingent on observation effort; while controlled statistically, unobserved behaviours are not confirmed absences.
• Habitat classification into forest vs savannah simplifies a continuum, potentially underestimating environmental gradients.
• Non-tool behaviours were underrepresented, increasing uncertainty in subset analyses.
• No direct evidence linking specific behaviours to fitness outcomes; adaptive interpretations are inferred from energetics or plausibility.
• Historical refugia designations and distances are proxies that may not capture the full complexity of past environments.
• Sampling limitations at smaller spatial scales and incomplete genome–behaviour association knowledge constrain broader generalizations.