Bottlenose dolphins (Tursiops truncatus) produce a wide variety of sounds but little is known about the function and organization of their vocal repertoires. This thesis investigates several aspects of call usage and compares the biological validity of classification methods for dolphin whistles. Passive acoustic localisation methods were used to identify which animal produced a sound. Observations of captive dolphins in the Zoo Duisburg, Germany, showed that signature whistles are almost only used when the group was split up, but not if all animals swam in together in the same pool. This finding supported the hypothesis that signature whistles are cohesion calls. Whistles from these observations were used to compare whistle classification conducted by eye with three computer methods using different similarity measures. Only the human observer classification was able to recognize whistle types that were used in a context- specific way by the animals confirming the power of this common classification method. Copying of signature whistles and whistle matching between animals was rare in captivity. However, observations of whistle interactions in the Moray Firth, Scotland, showed that wild dolphins do not tend to interact vocally in general, but that whistle matching was more frequent than expected by chance. Whistle matching in captivity was rare. Sound pressure measurements of dolphin whistles in the wild showed that source levels can reach up to 169 dB re 1 ?Pa and that the active space of a dolphin whistle can range up to 38 km. Finally, observations of foraging in wild dolphins revealed that they produce low frequency braying sounds in this context. Other dolphins would rapidly approach the caller in response to a bray. However, it is not clear whether brays function to attract conspecifics or manipulate prey behaviour.