More Than Just Kibbles: Keeper Familiarity and Food Can Affect Bonobo Behavior

Captive animal welfare is influenced by internal states and external factors such as environment, food, and human caregivers. While housing conditions have been widely studied, the roles of keepers (and their familiarity to animals) and food quality/availability are increasingly recognized as important, particularly for cognitively and emotionally complex great apes like bonobos. Prior research shows keeper-animal interactions can alter fearfulness and social behaviors, and food availability/quality can modulate arousal, anxiety, affiliation, and aggression. This study investigates how keeper presence/familiarity and food presence/quality affect bonobo behaviors expressing positive and negative emotional valence (play, aggression, anxiety), requesting gestures, and affiliative sociosexual interactions. The authors formulated predictions: (P1) anxiety would decrease with keeper presence and be higher without food; (P2) sociosexual interactions would decrease with keeper presence and increase with food, especially high-quality; (P3) requesting gestures would increase with keeper presence, especially with more familiar keepers, and be higher when high-quality food was present; (P4) play would increase with keeper presence/familiarity and be higher without food; (P5) agonism would be unaffected by keeper variables but increase with high-quality food.

Evidence across mammals and primates indicates keeper-animal relationships can function as enrichment, modulating fearfulness and affiliation, with animals discriminating among keepers and their familiarity. Food restriction or timing can induce anticipation-related anxiety (e.g., scratching, yawning, stress vocalizations) and increase conflicts, while high-quality food can either enhance affiliation (including play) or elevate competition and anxiety depending on species. Bonobos are generally tolerant with low aggression, but aggression can increase around valued food. Sociosexual interactions in bonobos serve tension reduction and bonding, often increasing in tense contexts (including post-conflict) and around feeding. Requesting gestures are intentional communicative acts used to obtain resources and are influenced by recipient familiarity. This background motivates testing interactive effects of keeper features and food type on bonobo social and emotional behaviors.

Study site and subjects: A stable group of 17 bonobos (Pan paniscus; age range 2–51 years; adults: 3 males, 6 females; juveniles: 2 males, 2 females; infants: 1 male, 2 females; one 2-year-old weaning female) housed at La Vallée des Singes (Romagne, France) with access to an indoor enclosure (~500 m²) and a wooded outdoor island (~1 ha). Animals typically moved freely between areas from ~09:30 to ~18:30 except in bad weather. Group management rotated among four keepers (2 men, 2 women); two were categorized as more familiar (≥5 years, ≥5 days/week) and two as less familiar (occasional, <5 years, 1–2 days/week). Feeding occurred four times daily (11:30, 14:30, 15:45, 17:00).

Data collection: July–September 2019 (~10:00–18:00) when bonobos were outdoors. Behaviors were video-recorded and coded using all-occurrences sampling. Target categories: anxiety/arousal behaviors (scratching, yawning), sociosexual interactions, play, agonistic behavior, and requesting gestures (operational definitions referenced in Table S2). Coding was by a trained observer with 100% agreement with the trainer before analysis.

Experimental conditions and sampling: Randomly selected two-minute time slots under three conditions: (1) keeper-present/food-unavailable (keeper visible with basket just before or during distribution when waiting for next ration); (2) keeper-present/food-available; (3) keeper-absent/food-unavailable. Keeper-absent/food-available did not occur in view/time window. Total analyzed: 118 video sequences (236 minutes), balanced across conditions and keeper/food features.

Food categories: (a) absence of food; (b) low-quality, large pieces (fruits/vegetables; ~23 kcal/100 g); (c) high-quality, small kibbles (St Laurent, Old World primate kibbles; ~370 kcal/100 g). Keepers threw food pieces to animals; fruits/vegetables were catchable in small numbers; kibbles were many small items dispersed.

Keeper familiarity: Two keepers (1 man, 1 woman) labeled more familiar; two (1 man, 1 woman) labeled less familiar.

Statistical analysis: Count data distribution compared via BIC for Poisson vs negative binomial; negative binomial selected. Generalized Linear Models (GLMs) with negative binomial errors (R, MASS::glm.nb). First set (GLM a; N=84 for keeper presence vs absence) tested effects of keeper presence (0/1) and food type (0=none; 1=low; 2=high) for each target behavior (anxiety: GLM1a; sociosexual: GLM2a; requesting gestures: GLM3a; play: GLM4; agonism: GLM5). If keeper presence significant, a second model (GLM b; N=56) tested keeper sex and familiarity. If familiarity significant, a third model (GLM c; N=56) tested keeper identity. Model comparison via likelihood ratio tests; predictor p-values via drop1 LRT; significant multinomial predictors followed by Tukey post hoc (multcomp) with Bonferroni adjustments. Effect sizes reported (effectsize package).

• Anxiety-related behavior: Full model significant (χ²=45.835, df=3, p<0.001). Anxiety increased with keeper presence (p=0.002). Food effect significant (p<0.001): anxiety higher in absence of food and with low-quality food than with high-quality food (Tukey: low vs absence Est=−1.031, SE=0.187, Z=−5.525, p<0.001; high vs absence Est=−2.161, SE=0.292, Z=−7.405, p<0.001; high vs low Est=−1.130, SE=0.302, Z=−3.745, p<0.001). Keeper sex/familiarity had no effect (GLM1b not different from null, p=0.571).
• Sociosexual behavior: Full model significant (χ²=31.102, df=3, p<0.001). Increased with keeper presence (p<0.001) and varied by food type (p=0.007); highest when food absent (low-quality vs absence Est=−1.038, SE=0.387, Z=−2.679, p=0.020; high-quality vs absence Est=−0.309, SE=0.551, Z=−0.561, p=0.838). With keeper features, sociosexual behavior was higher with more familiar keepers (GLM2b full vs null χ²=6.586, df=2, p=0.037; familiarity p=0.007). Keeper identity model not significant overall (p=0.060).
• Requesting gestures: Full model significant (χ²=67.991, df=3, p<0.001). Increased with keeper presence (p=0.020) and food type (p<0.001), peaking with low-quality food (Tukey: low vs absence Est=3.150, SE=0.404, Z=7.796, p<0.001; high vs absence Est=0.981, SE=0.608, Z=1.614, p=0.234; high vs low Est=−2.169, SE=0.549, Z=−3.950, p<0.001). More familiar keepers elicited more gestures (GLM3b full vs null χ²=6.320, df=2, p=0.042; familiarity p=0.005). Keeper identity significant (GLM3c χ²=9.577, df=3, p=0.023); highest to the more familiar woman and man; less familiar keepers elicited fewer gestures (e.g., less familiar woman Est=−2.294, SE=0.696, Z=−3.297, p=0.001).
• Play: No significant effects (GLM4 full vs null χ²=5.452, df=3, p=0.142).
• Agonistic behavior: Full model significant (χ²=98.657, df=3, p<0.001). Keeper presence not significant (p=0.388). Food type strongly increased agonism (p<0.001), highest with high-quality food (Tukey: low vs absence Est=2.495, SE=0.228, Z=10.952, p<0.001; high vs absence Est=2.956, SE=0.271, Z=10.904, p<0.001; high vs low Est=0.461, SE=0.217, Z=2.123, p=0.084).

Findings show bonobo behaviors around feeding are shaped by both food and keeper factors. Anxiety rose when the keeper was present and when food was absent or of lower quality, consistent with food anticipation when keeper presence signals imminent provisioning. Sociosexual interactions increased with keeper presence and especially with more familiar keepers, and were highest when food was not yet available—consistent with a tension-reduction function to manage arousal associated with expected feeding and, potentially, stronger arousal when familiar keepers arrive. Requesting gestures, as intentional signals to obtain resources, increased with keeper presence and familiarity, and were most frequent with large, catchable fruit/vegetable pieces (where gestures can secure discrete items), but not with dispersed small kibbles, where gesturing is less advantageous. Play showed no short-term modulation by keeper or food in the brief two-minute windows around feeding, aligning with prior work suggesting play may vary over longer pre-feeding periods. Aggression was unaffected by keeper variables but increased with high-quality food, reflecting competition over highly valued resources and close proximity while searching for kibbles. Overall, results indicate welfare-relevant behaviors depend on the interplay of caregiver familiarity and food characteristics, underscoring the importance of including human-animal relationship variables in management strategies for great apes.

Keeper presence/familiarity and food presence/quality jointly influence bonobo behavior. More familiar keepers were associated with higher sociosexual interactions and more requesting gestures; high-quality food increased aggression; anxiety rose with keeper presence (likely via food anticipation) and when food was absent or low-quality; play was unaffected in the short observation windows. Management recommendations include: randomizing feeding times; minimizing latency between keeper arrival and food distribution and reducing pauses during feeding to limit arousal/anxiety; favoring constant availability of adequate-quality food to reduce aggression; and accounting for social bonds with keepers to enhance tension-buffering strategies. These results emphasize that optimizing great ape welfare requires considering more than food alone, integrating caregiver-related variables given great apes' human-like cognitive and emotional sensitivities. Future work should further explore additional keeper characteristics and broader contexts to refine species-specific management.

• Keeper presence was always associated with forthcoming food provisioning in this study, making it difficult to fully disentangle effects of keeper presence from food expectation.
• Observations were confined to brief two-minute slots around feeding (including short pauses), which may have limited detection of behaviors (e.g., play) that vary over longer pre-feeding periods.
• The condition keeper-absent/food-available did not occur within the observable enclosure/time window, precluding analysis of that combination.