Social and ecological drivers of behavioural diversity in wild Costa Rican capuchins

The extent of behavioural diversity versus synchrony exhibited among members of the same social group impacts group cohesion. Understanding social and environmental drivers of behavioural diversity has implications for understanding the evolution of group living and has been an enduring goal of biologists. Here we analyse a 15 years behavioural dataset from wild habituated capuchin monkeys to explore social and environmental drivers of behavioural diversity. Between 2007-2022, instantaneous scan samples were recorded every 30 minutes from 214 individuals across 8 social groups, ranging in size from 7-35 individuals. We examine social, ecological, and demographic predictors of behavioural diversity at the group and individual level, with a focus on the impact of group size. We find that the behavioural richness and diversity of behaviours were greater in large groups compared to small and medium-sized groups, that individuals in small groups exhibited higher diversity across the day compared to individuals in medium and large groups, and that daily maximum temperature and rainfall were associated with decreased behavioural diversity at the group level. Overall, group size appears to impact important aspects of behavioural diversity within a group and among individuals in this socially complex species. The breakdown in behavioural synchrony of large groups likely influences movement ecology and social relationships and may contribute to fission events that occur at upper limits of group size. In the present study, we leverage a large dataset to provide new insights into how social and environmental variables can influence behavioural dynamics in an intelligent, social mammal. 

Methods
Detailed account of methods are available in the corresponding manuscript. Please see this abridged version of the methods, modified from the manuscript: We collected behavioral data on eight socia; group in rotations of 2-5 consecutive days, from sunrise to sunset. Each group was followed for 1-3 such rotations per month. Inter-observer reliability (IOR) is tested during a training period for new researchers, and they begin to collect data only once their IOR is >95% with established researchers. We conducted instantaneous scan sampling every 30 minutes (Martin et al. 1993). For the majority of contact hours, multiple researchers (n = 2-5) were present for each scan. For each scan, we recorded the instantaneous state behavior for as many individuals as we could identify within a window of up to 10 minutes, using an established ethogram (Supplemental Table 1A). Within foraging behaviors, we collected more specific foraging types (Supplemental Table 1B) as well as the type or species of food (when possible) that each individual was foraging on. Prior to each scan, each member of the observer team would disperse to maximize the possibility of locating each group member as quickly as possible. One observer would record each data point, and other observers would dictate the behavior of the individual primates, to minimize duplicate entries. In the rare cases where duplicate entries were recorded, the second entry was removed prior to data analysis. Typically, the behaviors of the majority of group members were recorded within the first ca. 3 minutes of each scan. During times when the monkeys were widely dispersed (e.g., after an intergroup encounter or during inclement weather), observers communicated via two-way radio to help ensure that data points were recorded as close together as possible, but researchers would use the full 10-minute window to locate individuals. Not all monkeys in the group were always able to be found in each scan. We have taken steps to account for this in our models, as detailed below. Finally, to minimize bias in data collection, such as over-representation of more centrally located monkeys, we began sampling with different individuals each time, and moved our positioning within the group to equalize, as much as possible, the coverage and order of recording different group members. We additionally collected data on rainfall, temperature, and fruit biomass. These variables were z-scored in our analysis. All code to reproduce these results is available at https://github.com/webbshasta/behavioral_diversity