A cross-center comparison of the relationship between matriline fragmentation, grooming cohesion, and agonistic behavior in captive rhesus macaque (Macaca mulatta) social groups

Brianne A. Beisner; Brenda McCowan; Mollie A. Bloomsmith; Lauren Lacefield; Kelly F. Ethun

doi:10.1016/j.applanim.2026.107036

Applied Animal Behaviour ScienceMay 3, 2026Animal Behavior & CognitionPeer-reviewed journal article

A cross-center comparison of the relationship between matriline fragmentation, grooming cohesion, and agonistic behavior in captive rhesus macaque (Macaca mulatta) social groups

Brianne A. Beisner iD, Brenda McCowan iD, Mollie A. Bloomsmith iD, Lauren Lacefield, Kelly F. Ethun iD

PDF DOI Preprint

In-depth full-text digest available · Jump to digest ↓

Abstract

A major challenge in managing captive-bred rhesus macaque social groups is mitigating deleterious aggression before it escalates to social instability. Prior work at the California National Primate Research Center (CNPRC) showed that fragmentation of matrilineal structure--reflected in lower average kinship among female kin--is associated with weakened cohesion in grooming networks and higher rates of intense aggression. We tested the generality of these findings by analyzing data from 105 matrilines across 16 social groups at CNPRC and Emory NPRC (ENPRC), which differ in group size, natal male management, and housing. Using generalized linear models, we found that matrilines with lower mean kinship coefficients showed greater grooming fragmentation, even after accounting for network density. Threshold analyses identified a mean kinship of 0.16 as the point at which grooming cohesion declined most consistently across both centers, highlighting a biologically meaningful level of relatedness for maintaining kin-biased social bonds. Patterns of severe aggression differed by target and center: across both centers, matrilines with lower mean kinship directed proportionally more severe aggression toward kin. However, for aggression toward all group members, lower kinship predicted more severe aggression only at ENPRC; at CNPRC, this effect emerged only when natal male aggression was included. Our results demonstrate that mean matrilineal kinship is a robust indicator of family cohesion and latent social instability across management settings. Nepotistic threshold analysis provides a practical tool for managers to identify matrilines at risk for social fragmentation and implement interventions before intra-family aggression emerges.

In-depth summary

Claude Opus 5.5 · xhigh effort · 5 Nov 2026 · from the full text of version 1 (21 Jan 2026) · CC BY 4.0

Across 105 rhesus macaque matrilines at two primate centers, family grooming cohesion drops below a mean kinship threshold of 0.16, while lower kinship continuously increases severe within-family aggression.

Severe intrafamily aggression threatens the stability of captive rhesus macaque breeding groups, and prior evidence linking pedigree fragmentation to reduced grooming cohesion came from a single facility. Comparing 105 matrilines across two primate centers with distinct group sizes and natal male management tests whether kinship thresholds generalize across colonies.

Key results

ShownLower mean matrilineal kinship consistently predicted greater family grooming fragmentation across both primate centers.
Across 105 matrilines (47 CNPRC, 58 ENPRC; modularity 0.28-0.53, mean 2.23 ± 2.63 SD or 24.1% of females clustering outside their primary family community), lower mean kinship predicted more females clustering separately in full networks (β = -5.67, SE = 1.58, p < 0.001) and density-matched networks (β = -4.67, SE = 1.38, p < 0.001). · RESULTS; Mean kinship coefficient predicts matriline…
ShownMatrilineal grooming cohesion followed a discrete relatedness threshold near r = 0.16 rather than a purely continuous decline.
Binary kinship cutoff GLMs (testing r = 0.10 to 0.25) outperformed continuous kinship models, peaking at a cutoff of 0.16 in combined full networks (ΔAICc = -5.7; CNPRC ΔAICc = -7 at 0.16, ENPRC ΔAICc = -0.3 at 0.175) and in density-matched networks (combined ΔAICc = -8.1, CNPRC ΔAICc = -5.3, ENPRC ΔAICc = -3.1). · Abstract; Kinship Threshold Analysis; Grooming
ShownMatrilines with lower mean kinship directed a higher proportion of severe aggression toward female relatives across both centers.
In quasibinomial GLMs of kin-directed aggression (n = 103 matrilines), including matriline rank, mean kinship coefficient, and center outperformed the null model (F 3,101 = 9.96, p < 0.001), showing higher proportions of severe (biting/pinning) kin aggression in lower-kinship and higher-ranking matrilines. · Aggressive Behavior and Matriline Rank Variables; Proportion of Aggression Toward Kin That Was…
ShownSevere aggression toward all group members scaled continuously with lower kinship and depended on center-specific natal male presence.
Lower female kinship predicted higher severe aggression at ENPRC (p = 0.008) where natal males are removed by age 3, but at CNPRC only when natal male aggression was included (p = 0.006 vs. p = 0.36 or p = 0.09 for females only); continuous kinship outperformed all threshold models (Residual Deviance = 308.97 vs. 347.74-381.76). · Proportion of All Aggression That Was Severe; Aggression

Methods, data and reproducibility

The authors analyzed grooming scan samples and agonistic event records from 105 matrilines (≥4 adult females each) across 7 CNPRC groups and 9 ENPRC groups, building weighted female grooming networks in igraph alongside density-matched networks (density ≤0.20). Walktrap community detection and Perc dominance ranks were modeled with GLMs to test continuous and binary (r = 0.10-0.25) kinship thresholds.

Observational adult female grooming networks, agonistic interaction logs, and maternal pedigree kinship coefficients from CNPRC and ENPRC · rhesus macaque (Macaca mulatta) · n = 105 matrilines across 16 social groups (47 CNPRC, 58 ENPRC; 103 for kin model)

Methods & tools: R, RStudio, igraph, Walktrap algorithm, Perc, MASS, stats, lme4, HandBase
Reagents & environment: R version 3.6.1 and RStudio with packages igraph (Walktrap algorithm), Perc, MASS, stats, and lme4
Reproducibility gaps: Behavioral datasets and pedigree summary tables are only available from the corresponding author upon request, and analysis scripts are not publicly archived.

Limitations (3)

Group selection was non-random: ENPRC groups were partly drawn from baseline phases of a male introduction study, excluded groups without matrilineal structure, and oversampled larger groups. · stated by the authors
Grooming networks and kinship calculations included only adult females and maternal lineages, excluding paternal relatedness and evaluating natal male effects on cohesion only indirectly. · stated by the authors
Results from captive breeding enclosures may not generalize directly to free-ranging macaque populations, which can resolve intragroup tension through group fission. · stated by the authors

Read this if you study primate social networks, kin selection thresholds, or behavioral management of captive non-human primate breeding colonies.