Abstract As echolocators, bats are an ideal model to study sensory ecology because they actively adjust their signal to sensory constraints in their environment. The general interest behind the work presented in this thesis is to link the behaviour of bats to their adaptive signal design in order to clarify how bats integrate complex echoes in a foraging context. I also wish to bring new perspectives on how to study bat behaviour in a controlled laboratory environment, by presenting a novel apparatus developed to study directional hearing in bats. The model species is the nectar-feeding bat Glossophaga soricina because of its easy maintenance in laboratory setting. In chapter three, I tested the flight and echolocation behaviour of the nectar-feeding bats by stereo videogrammetry as they approached a rewarded feeder in various cluttered conditions. The bats adjusted their trajectory and speed to the clutter and target proximity. Their calls (multi-harmonic broadband FM) were shorter and separated with shorter interval as they approached their target. They also adjusted their call frequency bandwidth in clutter. I argue that bats exploited different aspects of their calls according to context (e.g. overlap avoidance, increase in resolution). In chapter four, I investigated the source levels of the bats by stereo videogrammetry when they approached a target in different clutter conditions. The bats lowered their source level as they approached the target and were found to be louder than previously though. I propose that they compensate source level to maintain constant received echo intensity. They also adjust their calls to louder source levels when flying through narrower corridors. This may benefit their perception of the target by increasing the conspicuousness of the echo in narrower corridors. In chapter five, I present novel behavioural data testing the minimum audible angles of the bats in a 2AFC (2 alternative forced choice) lateralisation task of phantom echoes during echolocation. The three bats tested had MAA (Minimum audible angle) of 7.8°, 10.8° and 23°. I hypothesise that some bats used an alternative strategy to the trained task. Increasing the motivation level of the bats could reveal more consistent performances. I also tested the bat performances when three different frequency bandwidth of the phantom echo were played back, in order to evaluate the spectral cues for directional information. The bats did not perform differently in the three treatments. I argue that the spectral cues are not essential for a lateralisation task but would be crucial in a vertical localisation task.