Observation methods in animal behaviour: a simulation study of performance

Alexander Mielke , Camille Testard, Alba Motes-Rodrigo, Lauren J. N. Brent, Delphine De Moor

Most behavioural studies rely on systematic samples of behaviour, as observing and recording all behavioural events that occur is rarely feasible. Choosing an observation method involves several key decisions, including which individuals to observe, how to sample their behaviour, and how to distribute sampling effort over time. These decisions influence how closely behavioural estimates reflect the true occurrence of behaviours and how comparable estimates are across studies using different methods. Here, we used a simulation approach to evaluate the performance of different observation methods in terms of how accurately and precisely they represent true behavioural occurrences across varying contexts. We simulated behaviours differing in duration, frequency, and observability, in animals living in groups of different sizes and terrains with varying visibility. We then tested how the two most common observation methods—focal follows and group scans—captured these behaviours across different study durations, scan intervals, and focal lengths. We found that focal follows generated more accurate and precise behavioural estimates for short, rare behaviours, while group scans performed better for longer, more common behaviours. Group scans also performed better in larger group sizes and shorter study durations, as long as a large proportion of individuals was visible. We provide researchers with an interactive tool, the SIMBO app, to explore which observation method might be best suited to the specific properties of their system and research question. Overall, our study and app offer quantitative guidance on the performance of focal follows and group scans across contexts and highlights potential pitfalls for comparative research using data collected with different methods.