Invited Speaker

Forensic Speaker Identification: Phonetic and Linguistic Perspectives
Paul Foulkes, University of York, UK

It is widely accepted that there is no vocal equivalent of a fingerprint or a DNA profile which can offer irrefutable proof of speaker identity. Voices vary as a function of many factors – not only in response to the biological endowment of the speaker but also in systematic ways related to the social, situational and stylistic setting of a speech act. Robust speaker characterisation therefore demands an understanding of the many factors that yield variability.

In this talk I aim to provide a critical survey of linguistic and phonetic research on variation in speech. Drawing on authentic case materials I’ll provide a taxonomy of sources of variation. I’ll address principles and also problems of linguistic-phonetic methods, but I will argue that they can make a crucial contribution to forensic speaker identification.

The discussion will be embedded in the broader context of how phonetic-linguistic evidence is handled in courts around the world, with a view to raising controversial issues for debate at the Odyssey meeting. Potential topics include the continuing reliance of some laboratories on ‘voiceprinting’, the use of likelihood ratios, and the Daubert criteria.