This study explores the potential of utilising certain prosodic qualities of function-specific vocal expressions in order to design effective non-speech user interface sounds. In an empirical setting, utterances with four context-situated communicative functions were produced by 20 participants. Time series of fundamental frequency (F0 ) and intensity were extracted from the utterances and analysed statistically. The results show that individual communicative functions have distinct prosodic characteristics that can be statistically modelled. By using the model, certain function-specific prosodic cues can be identified and, in turn, imitated in the design of communicative interface sounds for the corresponding communicative functions in human-computer interaction.