The ability of dolphins to produce pulse sounds with a bandwidth of up to 150 kHz, has been intensively studied with focus on their sonar capabilities, i.e. the detection, ranging and discrimination of objects of various shapes, textures and material. The sonar pulses have been found to be directional, with a -3db beam width of about 10 degrees in both the vertical and the horizontal plane, and with the beam axis ea. 5 degrees above that of the rostrum. Pulse sounds also frequently occur in social situations (Paper I), but studies of these sounds traditionally have been limited to their audio band, omni-directional components. Hypothesising that social pulse sounds in bottlenose dolphins (Tursiops truncatus) also were directional, early recordings within this study were done across a net barrier, in a narrow channel connecting two pools, to increase the probability to record such sounds with a fixed hydrophone. These recordings confirmed the presence of ultrasonic frequencies in pulse sounds emitted in aggressive interactions, where the interacting animals were oriented face-to-face on either side of the net barrier. There were also indications that they were directional (Paper I). These sounds also had a pronounced energy peak 1 sec), labelled Machine-gun sound, occurred frequently in the initial phase of the aggressive interactions. While emitting this sound, the animal kept its rostrum, and possibly the sound beam axis aimed at the agonist, suggesting that it was deliberately used to expose the opponent to powerful broadband pulse sounds, probably to provoke a fight (Paper III).To avoid the restrictions on the behaviour imposed on the dolphins by the net barrier setup, an acoustic tag system was developed (Paper II). The tag was attached to the receiving dolphin instead of the transmitter, and made it possible to study the use of the directionality of the pulse sounds in social contexts, while the interacting dolphins were fully free-swimming. The tag, based on click detectors, recorded the envelope of pulses within two narrow band pass frequency filters, centred on 70kHz and 120kHz, respectively (Paper II). The tag recordings verified the presence of energy within both filter frequency bands in social pulse sounds, in particular in aggressive interactions (Paper III). This thesis suggests that thedirectionality of such pulse sounds is used intentionally to address social signals to a specific individual, or group of individuals. The relatively loud omni-directional audio band component will facilitate for nearby conspecifics to monitor the progress of the interaction. It will also make the signal clearly audible to the opponents throughout an interaction, even if they may be tumbling around, being unable to keep the sound beam axis aimed at the other. Similar pulse bursts were recorded in association with what was interpreted as play-fight, between subadults and juveniles (Paper IV). However, unlike the real aggressive bursts, these bursts had a trailing FM whistle, and this combined signal is suggested to be an appeasement signal, analogous with the laugh and giggle observed in apes. It is supposed to prevent play-fights from escalating into physical and potentially dangerous real fights. It was never recorded in true aggressive interactions among adults.There were individual differences between the animals in amount of received directional pulse sounds while carrying the tag (Paper III). The socially most active of the females, having an intermediate position in the dominance hierarchy, received significantly more directional pulse sounds than the other, socially less active, but higher ranked female. The breeding male received the least number of pulse sounds. This lends support to the presence of echolocation manners, as suggested by Johnson and Norris (1994), making it socially unacceptable to aiming the sonar beam towards conspecifics unless the intention is to conveying a social message.Dolphins lack the facial plasticity playing a dominant role in the social life of many terrestrial mammals, e.g. primates, canids and felids. This directional signal modality is used to express internal emotions and can be addressed to selected individuals, further stressed by a direct glare. Dolphins completely lack such a facial plasticity but most likely not the need for expressing emotions. This thesis suggests that in dolphins the facial expressions may have been replaced bydirectional pulse sounds. The facial expressions in e.g., wolves, is supposed to be the result of a mixture of dominance/aggression vs. submission/fear emotions. We propose that the pulse sounds, seen in the aggressive interactions, reflect a similar mix of feelings. Obvious parameters in such a graded social signal system are pulse repetition rate, amplitude, duration and power spectrum. To study this further a broadband acoustic tag, preferably attached to several animals simultaneously, is required.