The Individual Lay Listener as a Variable in Speaker Individualisation Tasks

In some legal cases investigators rely on the testimony of witnesses who have heard rather than seen a crime in order to establish the identity of the perpetrator. These ‘earwitnesses’ may be invited to take part in a voice parade (VP). In contrast to their well-established visual counterpart, however, VPs are rare, being difficult to design, implement and interpret.

Most recent attempts to improve earwitness reliability have focused on the VP procedure itself, by finding optimal settings for variables that can be controlled by an investigator, such as the quality, duration, and presentation of the stimuli. While such approaches help establish confidence in the procedure, they cannot establish confidence in the individual witness, whose general ability to correctly recognise voices may be questioned in court. The present study therefore takes a different approach by characterising the individual witness as a variable in the identification procedure.

Two hundred and fifty phonetically untrained listeners took part in three psychometric voice processing tests of increasing difficulty: An unfamiliar voice discrimination test (N = 100), an unfamiliar voice recognition test in which participants were aware of the task (N = 100), as well as an unfamiliar voice recognition test in which participants were unaware of the task (N = 50).

The tests were capable of discriminating between a wide range of performances, including a total of four ‘super recognisers’, who markedly outperformed other participants. Results are discussed in relation to the relevant linguistic, psychological, and legal literature. Moreover, the feasibility of a screening test for earwitnesses is discussed that may provide a trier of fact with nuanced information about an individual witness’s ability, thus helping assess the weight of VP evidence.