Search for: All records

Animal behavior can be decomposed into a sequence of discrete activity bouts over time. Analyzing the statistical structure of such behavioral sequences can provide insights into the drivers of behavioral decisions. Laboratory studies, predominantly in invertebrates, have suggested that behavioral sequences exhibit multiple timescales and long-range memory, but whether these results can be generalized to other taxa and to animals in natural settings remains unclear. By analyzing accelerometer-inferred predictions of behavioral states in three species of social mammals (meerkats, white-nosed coatis, and spotted hyenas) in the wild, we found surprisingly consistent structuring of behavioral sequences across all behavioral states, all individuals, and all study species. Behavioral bouts were characterized by decreasing hazard functions, wherein the longer a behavioral bout had progressed, the less likely it was to end within the next instant. The predictability of an animal’s future behavioral state as a function of its present state always decreased as a truncated power-law for predictions made farther into the future, with very similar estimates for the power law exponent across all species. Finally, the distributions of bout durations were also heavy-tailed. Why such shared structural principles emerge remains unknown, and we explore multiple plausible explanations, including environmental nonstationarity, behavioral self-reinforcement, and the hierarchical nature of behavior. The existence of highly consistent patterns in behavioral sequences across our study species suggests that these phenomena could be widespread in nature, and points to the existence of fundamental properties of behavioral dynamics that could drive such convergent patterns. 

Abstract Individual differences in behavior are the raw material upon which natural selection acts, but despite increasing recognition of the value of considering individual differences in the behavior of wild animals to test evolutionary hypotheses, this approach has only recently become popular for testing cognitive abilities. In order for the intraspecific approach with wild animals to be useful for testing evolutionary hypotheses about cognition, researchers must provide evidence that measures of cognitive ability obtained from wild subjects reflect stable, general traits. Here, we used a multi-access box paradigm to investigate the intra-individual reliability of innovative problem-solving ability across time and contexts in wild spotted hyenas (Crocuta crocuta). We also asked whether estimates of reliability were affected by factors such as age-sex class, the length of the interval between tests, or the number of times subjects were tested. We found significant contextual and temporal reliability for problem-solving. However, problem-solving was not reliable for adult subjects, when trials were separated by more than 17 days, or when fewer than seven trials were conducted per subject. In general, the estimates of reliability for problem-solving were comparable to estimates from the literature for other animal behaviors, which suggests that problem-solving is a stable, general trait in wild spotted hyenas. 

Abstract Integrated community models—an emerging framework in which multiple data sources for multiple species are analyzed simultaneously—offer opportunities to expand inferences beyond the single‐species and single‐data‐source approaches common in ecology. We developed a novel integrated community model that combines distance sampling and single‐visit count data; within the model, information is shared among data sources (via a joint likelihood) and species (via a random‐effects structure) to estimate abundance patterns across a community. Parameters relating to abundance are shared between data sources, and the model can specify either shared or separate observation processes for each data source. Simulations demonstrated that the model provided unbiased estimates of abundance and detection parameters even when detection probabilities varied between the data types. The integrated community model also provided more accurate and more precise parameter estimates than alternative single‐species and single‐data‐source models in many instances. We applied the model to a community of 11 herbivore species in the Masai Mara National Reserve, Kenya, and found considerable interspecific variation in response to local wildlife management practices: Five species showed higher abundances in a region with passive conservation enforcement (median across species: 4.5× higher), three species showed higher abundances in a region with active conservation enforcement (median: 3.9× higher), and the remaining three species showed no abundance differences between the two regions. Furthermore, the community average of abundance was slightly higher in the region with active conservation enforcement but not definitively so (posterior mean: higher by 0.20 animals; 95% credible interval: 1.43 fewer animals, 1.86 more animals). Our integrated community modeling framework has the potential to expand the scope of inference over space, time, and levels of biological organization, but practitioners should carefully evaluate whether model assumptions are met in their systems and whether data integration is valuable for their applications. 

Abstract How social development in early‐life affects fitness remains poorly understood.Though there is growing evidence that early‐life relationships can affect fitness, little research has investigated how social positions develop or whether there are particularly important periods for social position development in an animal's life history. In long‐lived species in particular, understanding the lasting consequences of early‐life social environments requires detailed, long‐term datasets.Here we used a 25‐year dataset to test whether social positions held during early development predicted adult fitness. Specifically, we quantified social position using three social network metrics: degree, strength and betweenness. We determined the social position of each individual in three types of networks during each of three stages of ontogeny to test whether they predict annual reproductive success (ARS) or longevity among adult female spotted hyenasCrocuta crocuta.The social positions occupied by juvenile hyenas did predict their fitness, but the effects of social position on fitness measures differed between stages of early development. Network metrics when individuals were young adults better predicted ARS, but network metrics for younger animals, particularly when youngsters were confined to the communal den, better predicted longevity than did metrics assessed during other stages of development.Our study shows how multiple types of social bonds formed during multiple stages of social development predict lifetime fitness outcomes. We suggest that social bonds formed during specific phases of development may be more important than others when considering fitness outcomes.