Not deep just average: Improving the usability of lay-listener voice descriptions

Earwitness testimony from lay listeners can provide vital evidence in a criminal investigation. However, some research suggests that voice descriptions can be vague. This paper presents the findings of three experiments investigating lay-listener voice descriptions, with a view to developing a procedure for eliciting accurate and admissible earwitness testimony. Experiment 1 examined free voice descriptions provided by native English-speaking lay listeners (N=20). As predicted, lay-listeners frequently used vague, or subjective voice descriptions. Phonetic features were inconsistently described across voices, possibly related to a lack of confidence in describing (or perceiving) the same phonetic feature across different voices. Experiment 2 sought to improve the accuracy, consistence and detail of these voice descriptions. Participants (N=30) were asked additional questions about the pitch and tempo of each voice. Responses were given either as free descriptions, or using a rating scale. Results show that listeners in both conditions produced mostly comparable descriptions of these specific features, but varied when describing other voice features. Experiment 3 focused on the memory retention of vocal information from unfamiliar voices. Respondents listened to a single, unfamiliar voice. Following a simple word-based task they used a pilot protocol to record a description of the target voice. Preliminary results show that the protocol increases the amount of objective descriptors, and reduces the number of subjective descriptors as well as the time taken to provide a voice description. These results will contribute toward the design of an earwitness protocol to be piloted in the East Midlands.