Abstract

In certain crimes, a perpetrator’s voice may have been heard at the scene of a crime, but not recorded. In such cases, if the police have already located a suspect, a phonetician may be asked to construct a ‘voice parade’ to be used in earwitness evidence. Analogous to a visual identity parade, the witness is asked whether he or she can pick out the voice of the speaker heard at the crime scene from a line-up of recordings which includes the suspect’s voice and a number of foil voices. One of the most difficult aspects of constructing a voice parade is selecting the foil voices. As well as the many practical challenges this presents, there is currently no general theory of what makes voices sound similar which would enable foil voices to be chosen in a principled way. This talk will firstly present a general background to the construction of voice parades and their associated theoretical and practical issues. Secondly, an experimental investigation of the phonetic underpinnings of perceived voice similarity for Standard Southern British and York Englishes will be described, considering the roles played by aspects of speech such as fundamental frequency, formant frequencies, voice quality and articulation rate, alongside the effect of the listeners’ own accent background. Implications of the findings for voice parade construction will be discussed.