Shy or bold, all get caught: Two active capture methods show no behavioural bias in a large herbivore

Abstract

Behavioural studies often assume that sampled individuals represent the broader population, yet behaviour-dependent capture bias can distort behavioural and ecological inferences. While more risk-prone individuals are likely to be overrepresented in captured samples, the potential for bias in wild animals, particularly large mammals, remains unclear due to the inherent difficulties in assessing it. This study tested whether chemical immobilisation and drop nets yielded samples of Northern chamois ( Rupicapra rupicapra ) representative of the population with respect to ecologically relevant behaviours related to the individual exposure to risks. We recorded the spontaneously exhibited behaviours of 25 free-ranging adult chamois previously captured and tagged, alongside their non-captured counterparts. First, we assessed the repeatability of key behaviours with expected positive (eating, walking, aggression) or negative (standing, resting, group centrality) associations with risk exposure in captured chamois, to evaluate their individual consistency and suitability as indicators of inter-individual behavioural variation. Eating, standing, resting, and group centrality were significantly repeatable, while walking and aggression were not. Second, we compared the frequency of these repeatable behaviours across chamois being captured using either method and their non-captured counterparts observed in the same sessions, controlling for demographic and environmental confounding factors. Both frequentist likelihood ratio tests and Bayesian analyses provided strong evidence against the presence of capture bias, indicating that both methods yielded a behaviourally representative sample in terms of eating, standing, resting, and group centrality. With this first explicit assessment of capture bias in a large mammal, we challenge the assumption that capture systematically skews behavioural data in wild animals, demonstrating that chemical immobilisation and drop nets can provide unbiased samples of behavioural diversity in chamois, and possibly in other large herbivores. Given the importance of minimising sampling biases in ecological research, we highlight the good opportunity to validate capture methods before extrapolating behavioural and ecological inferences.